JPH10507118A - Signal processing device - Google Patents

Abstract

SUMMARY The present invention relates to a method and apparatus for analyzing two measurement signals received by a detector (20, 320) modeled to include a primary part and a secondary part. According to the model defined according to the invention, the coefficients relate to the two measurement signals. In one embodiment, the present invention involves using a transform to find the appropriate coefficients. In another embodiment, the invention uses statistical functions or Fourier transforms and window techniques to determine coefficients associated with the two measurement signals. The use of the present invention has been described in particular detail with respect to blood oximetry.

Description

DETAILED DESCRIPTION OF THE INVENTION                           Title of invention: Signal processing device                                 Background of the Invention Field of the invention   The invention relates to the field of signal processing. More specifically, the present invention provides a primary signal part and The processing of the measurement signal, including the second and When little is known about either the primary or secondary signal part Processing the measurement signal to remove or derive any of the minutes. Further More specifically, the present invention minimizes the correlation between the primary and secondary signal portions. Measured by a new method that simplifies the generation of primary and / or secondary signals Related to modeling signals. The present invention relates to physiological models including blood oxygen saturation systems. It is particularly useful for nitering systems.Description of related technology   Typically, the signal processor includes a composite measurement signal including a primary signal portion and a secondary signal portion. Used to remove or derive either the primary signal part or the secondary signal part from the signal Is done. For example, a composite signal may include noise and required parts. Secondary signal part is one If it occupies a different frequency spectrum than the next signal part, a low-pass filter, Uses traditional filtering techniques such as bandpass and high-pass filters To remove or derive either the primary or secondary signal portion from the overall signal can do. The primary signal part and / or the secondary signal part is present at a fixed frequency In some cases, fixed single or multi notch filters can also be used.   There should be an overlap in the frequency spectrum between the primary and secondary signal parts. Is a common case. To complicate matters further, the primary signal part and the secondary signal part The statistical properties of one or both of the minutes change over time. In such cases, Conventional filtering techniques are ineffective at extracting either primary or secondary signals. is there. However, the description of either the primary signal part or the secondary signal part can be derived If possible, use correlation cancellation such as adaptive noise cancellation to create the primary signal of the signal. Either the part or the secondary signal part can be removed and separated from the other parts. In other words, if there is sufficient information about one part of the signal part, the signal part Can be extracted.   Conventional correlation cancelers, such as adaptive noise cancellers, adapt to parts of the composite signal. To dynamically change its transfer function to remove it. However, correlation cancellation Can be a secondary reference that correlates only to the secondary signal portion, or a correlation only to the primary signal portion. Either a primary criterion is required. For example, if the measurement signal is If the noise criterion is available, the noise can be reduced by using a correlation canceller. Can be removed. This is common. Reference signal amplitudes correspond The amplitude of the primary or secondary signal portion need not be the same, but Are similar to the frequency spectrum of the primary or secondary signal part. With torr.   In most cases, nothing or most of the secondary signal portion and / or the primary signal portion Not known. These parts of the measurement signal, including the primary signal part and the secondary signal part One area in which information about can not be easily determined is physiological monitoring . Physiological monitoring generally involves measurement signals obtained from a physiological system, such as the human body. No. Measurements typically made using physiological monitoring systems are Electrograms, blood pressure, blood gas saturation (eg oxygen saturation), capnographs, other blood components Includes minute monitoring, heart rate, respiratory rate, electroencephalogram (EGG), and depth of anesthesia. other Types of measurement include measuring the pressure and volume of substances in the human body, such as cardiac output , Venous oxygen saturation, arterial oxygen saturation, bilirubin, total hemoglobin, respiratory test , Drug test, cholesterol test, glucose test, bleeding, and carbon dioxide test , Protein tests, carbon monoxide tests, and other in vivo measurements. Such measurement A complex problem that arises in the measurement is the external and internal movement of the patient during the measurement process (eg, Muscle movement, vein movement, and movement of a probe).   Known energy decay characteristics of selected forms of energy as it passes through the medium By using, many types of physiological measurements can be made.   Blood gas monitors are based on the measurement of energy attenuated by biological tissue or components. 1 is an example of a physiological monitoring system. Blood gas monitor is a test medium Light is sent through and the decay of the light is measured as a function of time. Reacts to arterial blood The output signal of the blood gas monitor includes a component that is a waveform representing the arterial pulsation of the patient. Suffering This type of signal, which contains components related to the person's beat, is called a plethysmographic wave. , Which is shown by the curve s in FIG. Plethysmograph waveform measures blood gas saturation Used in Like the heartbeat, the blood volume in the arteries increases or decreases, thereby causing The energy decay increases and decreases as indicated by the period wave s.   Blood is used as a medium to transmit light energy for blood gas attenuation measurement. It is typical to use nearby flowing fingers, earlobes or other parts of the human body. FIG. Fingers schematically indicated by include skin, fat, bone, muscle, etc., each of which is substantially predictable and Constantly attenuates the energy incident on the finger. However, for example, the thickness of the finger When the part is pressed unexpectedly, the movement of the finger causes the energy attenuation to become incidental. You.   An example of a more realistic measurement waveform S is shown in FIG. This shows the effect of the behavior . The primary plethysmograph waveform s of the signal is a waveform representing a pulse, and the sawtooth waveform in FIG. Corresponding to the shape pattern. Large displacement of signal amplitude caused by secondary operation The plethysmographic signal s becomes ambiguous. Even with small changes in amplitude, It is difficult to identify the primary signal component s in the presence of the signal component n.   Pulse oximeter is a blood gas monitor that non-invasively measures arterial oxygen saturation in blood Is one type. New oxygenated blood moves due to the pumping action of the heart Stimulated into the pulse, resulting in greater energy decay. Well understood in the technology As is known, arterial saturation of oxygenated blood is measured at different wavelengths. It can be measured from the valley depth relative to the peaks of the two plethysmographic waveforms. Figure As shown in the plethysmograph waveform shown in FIG. The motion artifact is added to the signal. Due to such motion artifacts, the measurement signal The issue is disturbed.                                 Summary of the Invention   The present invention relates to a U.S.A. named Signal Processing Apparatus. Patent application No. 08 / 132,812 (filed on Oct. 6, 1993) An improvement of the method and apparatus (the patent application is assigned to the assignee of the present application). Book The invention is directed to separating either the primary or secondary signal portion of a composite measurement signal. It includes several different embodiments that use the novel signal model according to the invention. 1 In one embodiment, the signal processor includes a first measurement signal and a first measurement signal. And obtaining a second measurement signal. The first signal comprises a first primary signal portion and a first secondary signal Including parts. The second signal includes a second primary signal portion and a second secondary signal portion. This These signals propagate energy through the medium and measure the attenuated signal after transmission or reflection. Can be obtained. Alternatively, these signals are the energy generated by the medium. It can also be obtained by measuring the energy.   In some embodiments, the first measurement signal and the second measurement signal are the first measurement signal or the second measurement signal. Processed to generate a secondary reference without the primary signal part from any of the measurement signals Is done. The secondary reference is a secondary signal portion of each of the first measurement signal and the second measurement signal. Correlates with The secondary criterion is passed through a correlation canceller such as an adaptive noise canceller. And is used to remove a secondary portion of each of the first measurement signal and the second measurement signal. You. The correlation canceller obtains a first input and a second input, and obtains all signals correlated with the second input. Is a device that removes the signal component from the first input. In this specification, Any unit that performs or nearly performs a function is considered a correlation canceller.   The adaptive correlation canceller dynamically changes its transfer function in response to the reference and measurement signals. Dynamic notch to remove from the measurement signal the frequency present in the reference signal by changing to It can be explained in analogy with a filter. Hence, a typical adaptive correlation canceller Receives the signal from which a certain component needs to be removed and receives the reference signal of the unnecessary portion. I believe. The output of the correlation canceller is used for the desired signal from which unnecessary parts have been removed. This is a valid approximation.   Alternatively, the first measurement signal and the second measurement signal may be a secondary signal from any of them. It can be processed to generate a primary reference without parts. The primary criterion is then Remove primary part of each signal of first measurement signal and second measurement signal via Seki canceller Can be used to The output of the correlation canceller is a second-order signal with the primary signal removed. A valid approximation to the signal that can be used for further processing on the same or auxiliary equipment. Can be used. In this capability, either the first measurement signal or the second measurement signal As a reference signal input to the second correlation canceller together with the By using the approximation, the first primary signal portion or the second primary signal respectively Any of the sign portions can be calculated.   Physiological monitors can benefit from the signal processor of the present invention. Living Often in a physical measurement, a first signal including a first primary part and a first secondary part. And a second signal comprising a second primary part and a second secondary part. Patient body Through the body (or a substance obtained from the human body, such as breath, blood, or tissue) To transfer energy into the vein or vein and measure the attenuated signal after transmission or reflection Thus, these signals can be obtained. Or, for example, in an electrocardiogram, The signal can be obtained by measuring the energy generated by the patient's body. Can be. Adaptive processing by processing the signal through the signal processor of the present invention Either secondary or primary reference input to correlation canceller such as noise canceller This is obtained.   One physiological monitoring device that would benefit from the present invention is a plethysmograph It is a monitoring system that measures a signal representing arterial pulsation called a wave. this The signal can be used for blood pressure measurement, blood component measurement, and the like. Of such use A particular example is pulse oximetry. Pulse oximetry measures oxygen saturation in the blood Including measurements. In such a configuration, the primary part of the signal is that blood flows near the skin Of arterial blood to energy attenuation as energy passes through a part of the human body Degrees. Due to the pumping action of the heart, the blood flow periodically increases and decreases in the arteries Periodic decay occurs, but the periodic waveform is a plethysmographic waveform representing arterial pulsation It is. The secondary part is noise. In the present invention, when energy passes through the human body The secondary part of this signal is related to the contribution of venous blood to the energy decay of Next, the measurement signal is modeled. The secondary part is also the Artifacts that can cause venous blood to flow unpredictably Thus, unpredictable attenuation occurs, and the periodic plethysmograph waveform is disturbed. Call Inhalation is also typically at a lower frequency than the patient's pulse rate, Causes a change in the noise portion. Thus, according to the present invention, a plethysmographic wave The measurement signal forming the shape, the primary part of which indicates the contribution of arterial blood to attenuation, The secondary part is modeled such that it is due to some other parameter.   A physiological monitor specifically applied to oximetry by pulse oximetry is , Two light emitting diodes that emit light of different wavelengths to generate a first signal and a second signal (LED). Each of the two different energy signals is, for example, a finger or an ear. After passing through an absorbing medium such as a glove, the attenuation of these two different energy signals is detected. The dispatcher records. The decay signal generally comprises a primary signal portion (attenuation for the artery) and Both parts of the next signal part (noise) are included. Static filter such as band pass filter The system often removes portions of the secondary signal outside the known bandwidth of interest. Concomitant or random generated by action or often difficult to remove The secondary signal portion remains with the primary signal portion.   A processor according to one embodiment of the invention removes a primary signal portion from a measurement signal. To obtain a secondary reference, which is a combination of the remaining secondary signal portions. The secondary criterion is Correlates to both secondary signal parts. The secondary reference and at least one measurement signal are A correlation key such as an adaptive noise canceller that removes random or concomitant parts of the signal. Entered into Jansera. This allows the primary pre-processor to be measured at one of the measurement signal wavelengths. A valid approximation to the chismographic signal is obtained. As known in the art, Quantitative measurement of oxygen-containing arterial blood volume in the human body is based on a variety of plethysmographic signals. Can be obtained in a way.   The processor of the present invention also removes the secondary signal portion from the measurement signal, and A primary reference, which is a combination of the next signal portions, can be obtained. This primary criterion is Correlates to both primary signal portions. At least one of the primary reference and the measurement signal is measured It is input to a correlation canceller that removes the primary part of the signal. This allows the measurement signal A valid approximation to the secondary signal at one of the wavelengths is obtained. This signal is Measuring venous blood oxygen saturation as well as removing secondary signals from assistive devices May also be useful.   According to the signal model of the present invention, each has a primary signal part and a secondary signal part The two measurement signals can be related by a coefficient. Defined by the present invention By associating the two equations with respect to the coefficients, the coefficients are calculated as arterial oxygen saturation and noise. Provide information about (venous oxygen saturation and other parameters). This aspect of the invention According to the model, the phase between the primary and secondary signal parts specified in the model The coefficients can be determined by minimizing the function. Therefore, the signal model of the present invention is It can be used in many ways to obtain information about the measurement signal. About this It will become more apparent in the detailed description of the preferred embodiments.   One aspect of the invention provides that at least two measurement signals S₁And S_TwoTo handle the This is the method used in the processor. Each of at least two measurement signals Includes a primary signal portion s and a secondary signal portion n, and the signal S₁And S_TwoHas the following relationship .                               S₁= S₁+ N₁                               S_Two= S_Two+ N_Two Where s₁, S_Two, N₁, N_TwoHas the following relationship:                 s₁= R_as_Two, And n₁= R_vn_Two In the above equation, r_a, R_vIs a coefficient.   The method of the present invention comprises a number of steps. s₁And n₁The agent that minimizes the correlation between Number r_aIs determined. Then, the determined r_aThe first signal and the second signal At least one of the signals is from at least one of the first measurement signal or the second measurement signal n is subtracted to form a clean signal.   In some embodiments, the clean signal is displayed on a display. In another embodiment, wherein the first signal and the second signal are physiological signals, the method further comprises: Processing the clean signal to obtain a physiological parameter from the first or second measurement signal. Determining a meter. In some embodiments, the parameter is arterial oxygen saturation. The degree of peace. In another embodiment, the parameter is an ECG signal. Of the measurement signal In another embodiment, wherein the portion shows a cardiac plethysmograph, the method comprises: Further, the method includes a step of calculating a pulse rate.   Another aspect of the invention involves a physiological monitor. The monitor has a primary part s₁And two Next part n₁Measurement signal S having₁Is configured to receive the first input. Mo Nita also has a primary part s_TwoAnd the secondary part n_TwoMeasurement signal S having_TwoReceive The second input is configured as follows. Advantageously, the first measurement signal S₁And the second measurement signal S_TwoHas the following relationship.                               S₁= S₁+ N₁                               S_Two= S_Two+ N_Two Where s₁And s_TwoAnd n₁And n_TwoHas the following relationship:                     s₁= R_as_Two, And n_l= R_vn_Two In the above equation, r_a, R_vIs a coefficient.   The monitor further has a scan reference processor, which scan reference processor_a In response to the multiple possible values of_aIs multiplied by each possible value of Each of the values is subtracted from the first measurement signal to obtain a plurality of output signals. Receives the first measurement signal A first input configured to communicate with the plurality of output signals from the saturation scan reference processor. And a correlation canceller configured with a second input to receive the plurality of output signals. Providing a plurality of output vectors corresponding to correlation cancellation between the first measurement signals. Integration configured with inputs to receive multiple output vectors from correlation canceller The means is responsive to the plurality of output vectors to determine a corresponding power for each output vector. Set. The extreme value detecting means has its input connected to the output of the integrating means. Extreme value detection Stage detects selected power in response to corresponding power for each output vector I do.   In some embodiments, the plurality of possible values is a plurality of possible values for the selected blood component. Corresponds to the value. In some embodiments, the selected blood component is arterial blood oxygen saturation. is there. In another embodiment, the selected blood component is venous blood oxygen saturation. Ma In yet another embodiment, the blood component selected is carbon monoxide.   Another aspect of the invention involves a physiological monitor. The monitor has a primary part s₁And secondary Minutes n₁Measurement signal S having₁Is configured to receive the first input. monitor Is also the primary part s_TwoAnd the secondary part n_TwoMeasurement signal S having_TwoTo receive A second input is configured. First measurement signal S₁And the second measurement signal S_TwoIs related to Obey.                               S₁= S₁+ N₁                               S_Two= S_Two+ N_Two s₁And s_TwoAnd n₁And n_TwoIs                       s₁= R_as_Two, N₁= R_vn_Two Where r_aAnd r_vIs a coefficient.   The conversion module is responsive to the first measurement signal and the second measurement signal, and_aof Respond to multiple possible values. The extreme value calculation module comprises at least one power curve R that minimizes the correlation between s and n according to_aAnd choose this r_aCorresponding from the value of Is calculated as an output. The display module displays the calculated saturation. The saturation value is displayed in response to the output of the sum value.                               BRIEF DESCRIPTION OF THE FIGURES   FIG. 1 is a diagram showing an ideal plethysmograph waveform.   FIG. 2 schematically shows a typical finger.   FIG. 3 shows a plethysmograph waveform including a contingency signal portion generated by the operation. FIG.   FIG. 4a is a schematic diagram of a physiological monitor that calculates a primary physiological signal.   FIG. 4b is a schematic diagram of a physiological monitor that calculates a secondary signal.   FIG. 5a can be used in a physiological monitor that calculates the primary physiological signal FIG. 3 is a diagram illustrating an example of an adaptive noise canceller.   FIG. 5b illustrates a physiological monitor that calculates a secondary motion artifact signal. FIG. 3 is a diagram showing an example of an adaptive noise canceller that can be used in the present invention.   FIG. 5c shows the transfer function of the multi-notch filter.   FIG. 6a is a schematic diagram of an absorbent material containing N components.   FIG. 6b is another schematic diagram of an absorbent material, including one mixed layer. It contains N components.   FIG. 6c is another schematic diagram of the absorbent material, including two mixed layers. N components.   FIG. 7a illustrates a monitor for calculating primary and secondary signals in accordance with one aspect of the present invention. It is a schematic diagram.   FIG. 7b shows the signal coefficient r₁, R_Two... r_nIdeal correlation canceller as a function of FIG. 4 shows the energy or power output of the radiator, in this particular example r_Three= R_aAnd r₇= R_vIt is.   FIG. 7c shows the signal coefficient r₁, R_Two... r_nNon-ideal correlation cancellation as a function of FIG. 4 shows the energy or power output of the sera, in this particular example r_Three= R_aIn R₇= R_vIt is.   FIG. 8 shows a combined process estimator including a least squares grid type prediction unit and a regression filter. 2 is a schematic model of a step.   FIG. 8a includes QRD least squares lattice (LSL) predictor and regression filter It is a figure showing a schematic model of a joint process estimating means.   FIG. 9 shows the execution of the joint process estimating means modeled in FIG. 8 in software. 9 is a flowchart showing a subroutine for performing the operation.   FIG. 9a shows the combination process estimation means modeled in FIG. It is a flowchart which shows the subroutine for performing.   FIG. 10 shows a combined process having least-squares lattice type prediction means and two regression filters. 3 is a schematic model of a source estimating means.   FIG. 10a shows a result having a QRD least-squares lattice type prediction means and two regression filters. It is a schematic model of a combined process estimation means.   FIG. 11 illustrates an example of a physiological monitor in accordance with the teachings of one aspect of the present invention. .   FIG. 11a shows a low noise emitter current driver with a digital-to-analog converter. It is a figure showing an example of a moving device.   FIG. 12 shows a pre-analog signal conditioning circuit of the physiological monitor of FIG. FIG. 3 is a diagram illustrating a digital conversion circuit.   FIG. 13 is a diagram showing further details of the digital signal processing circuit of FIG.   FIG. 14 shows operations performed by the digital signal processing circuit of FIG. FIG. 4 shows further details of the.   FIG. 15 shows further details regarding the demodulation module of FIG.   FIG. 16 is a diagram showing further details regarding the decimation module of FIG. It is.   FIG. 17 is a more detailed block diagram of the operation of the statistics module of FIG. FIG.   FIG. 18 is a block diagram of the operation of one embodiment of the saturation conversion module of FIG. It is a lock figure.   FIG. 19 is a block diagram of the operation of the saturation calculation module of FIG. You.   FIG. 20 is a block diagram of the operation of the pulse rate calculation module of FIG. You.   FIG. 21 is a block diagram of the operation of the operation artifact suppression module of FIG. It is a lock figure.   FIG. 21a illustrates the operation of the motion artifact suppression module of FIG. It is another block diagram.   FIG. 22 is a diagram showing a saturation conversion curve according to the principle of the present invention.   FIG. 23 is a block diagram of another embodiment of the saturation conversion for obtaining the saturation value. You.   FIG. 24 is a diagram illustrating histogram saturation conversion according to another embodiment of FIG. You.   FIGS. 25A to 25C are diagrams showing another embodiment for obtaining the degree of saturation.   FIG. 26 shows the secondary criterion n ′ (used in the processor and the correlation canceller of the present invention. t) or the red light wavelength λa = λ for determining the primary criterion s ′ (t)_red= 660nm FIG. 3 is a diagram showing signals measured in the first embodiment. The measurement signal has a primary part s_λa(T) and Secondary part n_λa(T).   FIG. 27 shows the secondary criterion n ′ (used in the processor and the correlation canceller of the present invention. t) or the infrared wavelength λb = λ for determining the primary criterion s ′ (t)_IR= 910 nm FIG. 4 is a diagram showing a signal measured during the measurement. The measurement signal has a primary part s_λb(T) and two Next part n_λb(T).   FIG. 28 shows the secondary criterion n ′ (t) determined by the processor of the present invention.   FIG. 29 shows λa estimated by correlation cancellation using the secondary criterion n ′ (t). = Λ_red= Signal S measured at 660 nm_λaPrimary part s of (t)_λa(T) A valid approximation to_λaIt is a figure showing (t).   FIG. 30 shows λb estimated by correlation cancellation using the secondary criterion n ′ (t). = Λ_IR= Signal S measured at 910 nm_λbPrimary part s of (t)_λb(T) A valid approximation to s "_λbIt is a figure showing (t).   FIG.₁, S_TwoAnd S_ThreeFor obtaining an electroencephalogram (ECG) signal indicated by FIG. 3 shows a set of three concentric electrodes, ie, a three-electrode sensor. Each EGG signal is Includes primary and secondary components.   FIG. 32 is a diagram showing a tap delay line FIR filter.   FIG. 33 is a diagram showing a vector sum of FIR filter outputs for two frequencies. is there.   FIG. 34 is a diagram illustrating an M-th all-pass subfilter.   FIG. 35 shows complementary low-pass and high-pass filters from a two-pass all-pass network. It is a figure showing a filter.   FIG. 36 shows Hilbert as a converted half-band filter. It is a figure showing a conversion all pass filter.   FIG. 37 is a diagram showing a resampling two-pass all-pass filter.   FIG.^-FourFIG. 8 is a diagram showing a spectrum of an iterative two-pass filter having the following polynomial: is there.   Fig. 39 shows a filter bar concentrated in the quadrant of four channels and concentrated in the main axis direction. It is a figure showing the spectrum of a link.   FIG. 40 is a diagram illustrating the low-pass conversion of the all-pass filter.   FIG. 41A is a diagram showing an all-pass filter structure.   FIG. 41B is a diagram showing a multi-phase all-pass structure.   FIG. 42 is a diagram illustrating a pole-zero distribution of the M-order all-pass subfilter.   FIGS. 43A-43D show typical phase responses of the all-pass filter path and corresponding It is a figure showing an amplitude response.   44A and 44B show pole zeros for 2-pass and 5-pass polyphase all-pass networks. It is a figure which shows a plot and an amplitude response.   FIG. 45 is a diagram illustrating bandpass conversion of an all-pass filter.                               Detailed description of the invention   The present invention comprises a first measurement signal and a second measurement signal, each having a primary signal portion and a secondary signal portion. Includes systems that use measurement signals. In other words, the first composite signal S₁(t) = s₁( t) + n₁(t), the second composite signal S_Two(t) = s_Two(t) + n_Two(t), the system of the present invention Using the stem, either the primary signal portion s (t) or the secondary signal portion n (t) They can be separated. The output of the processed system is the secondary signal portion n (t ), Or a valid approximation to the primary signal part s (t) s "(t).   The system of the present invention provides that the primary and / or secondary signal portion n (t) is It may be necessary to include one or more parts such as measurable parts, incidental parts, and random parts. Useful for The primary signal approximation s "(t) or the secondary signal approximation n" (t) is the composite signal S ( t) as many secondary signal parts n (t) or primary signal parts s (t) as possible Obtained by leaving. The remaining signals are respectively the primary signal approximation s "(t) or Form any of the secondary signal approximations n "(t). A fixed portion of the secondary signal n (t) and Predictable parts include, for example, simple subtraction, low-pass, band-pass, and high-pass filters. It is easily removed using conventional filtering techniques such as tarring. Incidental part Are difficult to remove due to their unpredictable nature. Statistical and concomitant signals If some knowledge of At least partially from the measurement signal. However, the secondary signal n Information about the incidental part of (t) is often unknown. In this case, Filtering techniques are usually inadequate.   To remove the secondary signal n (t), the signal model according to the present invention uses the first measurement signal S₁ And the second measurement signal S_TwoIs defined as follows. S₁= S₁+ N₁ S_Two= S_Two+ N_Two s₁= R_as_Two, And n₁= R_vn_Two Or Where s₁And n₁Are at least somewhat (preferably substantially) correlated S_TwoAnd n_TwoAre at least somewhat (preferably substantially) uncorrelated. First measurement signal S₁And the second measurement signal S_TwoIs the correlation coefficient r as defined above_aas well as r_vAre related by The use and selection of these factors are described below. In more detail.   In one aspect of the invention, the signal model is a correlation canceller such as an adaptive noise canceller. Used in combination with Serra to remove or derive incidental portions of the measurement signal.   In general, a correlation canceller has two signal inputs and one output. These inputs One of which is either the secondary criterion n ′ (t) or the primary criterion s ′ (t), Are respectively the secondary signal part n (t) and the primary signal present in the composite signal S (t). Signal portion s (t). The other input is for the composite signal S (t) . Ideally, the output s "(t) or n" (t) of the correlation canceller is the primary signal It only corresponds to the part s (t) or the secondary signal part n (t). Often, correlation scans The most difficult task in the Serra application is the secondary part of the measurement signal S (t) The reference signals n ′ (t) and s ′ (correlating to the minute n (t) and the primary part s (t) respectively t) because, as described above, these parts This is because it is extremely difficult to separate from S (t). Signal processor of the present invention , The secondary criterion n ′ (t) or the primary criterion s ′ (t) is two different wavelengths λa and λb Are determined at the same time from two composite signals measured simultaneously or substantially simultaneously.   A general monitor block comprising a signal processor and a correlation canceller according to the invention. The diagrams are shown in FIGS. 4a and 4b. Two measurement signals S_λa(T) and S_λb(T) is Received by detector 20. To make some physiological measurements, one or more Those skilled in the art will appreciate that a dispenser is better. Each signal is signal adjusted It is adjusted by means 22a and 22b. Adjustments are signaled to remove certain parts. Procedures such as filtering the signal and amplifying the signal for ease of operation. However, the present invention is not limited to this. The signal is then converted to an analog-to-digital converter 24a and The data is converted into digital data by 24b. First measurement signal S_λa(T) is a book S in the description_λaA first primary signal portion, labeled (t), and n_λa(T) And a first secondary signal portion labeled. Second measurement signal S_λb(T) is , The first measurement signal S_λa(T), and is at least partially related to s_λbA second primary signal portion, labeled (t), and n_λbWith label (t) A second secondary signal portion to be transmitted. Typically, the first secondary signal portion n_λa (T) and the second secondary signal part n_λb(T) is the primary signal part s_λa(T) And s_λbUncorrelated and / or adjoined to (t). Secondary signal section Minutes n_λa(T) and n_λb(T) is often associated with patient movement in physiological measurements It arises from.   Signal S_λa(T) and S_λb(T) is input to the reference processor 26. Standard Processor 26 has a factor r_a= S_λa(T) / s_λb(T) or factor r_v= N_λa(T) / n_λb(T) or any of the second measurement signals S_λbRaised to (t) The second measurement signal S_λb(T) is the first measurement signal S_λaSubtract from (t). signal Coefficient factor r_aAnd r_vIs the two signals S_λa(T) and S_λb(T) was subtracted Sometimes the primary signal portion s_λa(T) and s_λb(T) or secondary signal part n_{λ a} (T) and n_λbIt is determined to erase one of (t). Therefore, the criteria The output of the processor 26 is the secondary signal portion n_λa(T) and n_λb(T) The secondary reference signal n ′ (t) = n in FIG._λa(T) -r_an_λb(T) or one Next signal part s_λa(T) and s_λb(T) the secondary reference signal of FIG. s' (t) = s_λa(T) -r_vs_λb(T). Reference signal n '(t ) Or s ′ (t) is the measurement signal S_λa(T) or S_λb(T) with one of The correlation canceller 27 inputs the reference signal n ′ (t) or s ′ (t) to the canceller 27. Using the secondary signal part n_λa(T) or n_λb(T) or primary signal part s_{λ a} (T) or s_λb(T) one of the measurement signals S_λa(T) or S_λb( Remove from t). The output of the correlation canceller 27 is a valid primary signal approximation s "(t) Or the secondary signal approximation n "(t). In some embodiments, the approximation s" (t) or n "(t). Are displayed on the display 28.   In one embodiment, the adaptive noise canceller 30 (an example of which is shown in FIG. ) Is used as a correlation canceller 27, and an optional secondary signal portion n_λa (T) and n_λb(T) is applied to the first signal S_λa(T) and second measurement Fixed S_λbRemove from (t). The adaptive noise canceller 30 of FIG. Part n_λa(T) and n_λbOne sample of the secondary criterion n '(t) correlated to (t) Have as input. The secondary criterion n '(t) is, as described in the text, the process of the present invention. The two measurement signals S_λa(T) and S_λb(T) You. The second input to the adaptive noise canceller is the first composite signal S_λa(T) = s_λa (T) + n_λa(T) or the second composite signal S_λb(T) = s_λb(T) + n_λb( t) is any sample.   Using the adaptive noise canceller 30 of FIG._λa(T) And the second measurement signal S_λbFrom (t) the primary signal part s_λa(T) and s_λb(T) Either one can be removed. The adaptive noise canceller 30 converts the primary signal Part s_λa(T) and s_λbOne sample of the primary criterion s ′ (t) that correlates to (t) Have as input. The primary criterion s' (t) is, as described in the text, Two measurement signals S by the sensor 26_λa(T) and S_λb(T). The second input to the adaptive noise canceller 30 is the first measurement signal S_λa(T) = s_λa (T) + n_λa(T) or S_λb(T) = s_λb(T) + n_λbAny of (t) This is a sample.   The adaptive noise canceller 30 uses the reference signal n ′ (t) or s ′ (t) and the measurement signal S_λa (T) or S_λb(T) functions to remove frequencies common to both. Base The quasi-signal is the secondary signal part n_λa(T) and n_λb(T) or primary signal part s_λa( t) and s_λb(T), the reference signal is It is incidental or behaves the same. The adaptive noise canceller 30 receives the reference signal Similar to dynamic multi-notch filter based on n '(t) or s' (t) spectral distribution Act to be able to   FIG. 5c shows a transfer function of an example of a multi-notch filter. Transfer function amplitude Notches or dips in the signal are attenuated or removed when the signal passes through the notch filter. Indicates the frequency to be dropped. The output of the notch filter excludes the frequency at which the notch exists. This is the composite signal that was dropped. Notch similar to adaptive noise canceller 30 The frequency to be changed continuously changes based on the input to the adaptive noise canceller 30.   The adaptive noise canceller 30 (FIGS. 5a and 5b)_λa(T), s "_λb (T), n "_λa(T) or n "_λbGenerate an output signal labeled (t) This output signal is fed to an internal processor 32 in the adaptive noise canceller 30. Will be back. The internal processor 32 executes its own transmission according to a predetermined algorithm. By adjusting the transfer function, b in FIG._λLabeled (t) and FIG. b then c_λThe output of the internal processor 32, labeled (t), is n_λa(T) or n_λb(T) or primary signal part s_λa(T) or s_λb( t). Output b of internal processor 32 of FIG. 5a_λ (T) is the measurement signal S_λa(T) or S_λbWhen subtracted from (t), the output signal s "_λa(T) = s_λa(T) + n_λa(T) -b_λa(T) or output signal s "_λb = S_λb(T) + n_λb(T) -b_λb(T) is obtained. The internal processor is "_λa(T) or s "_λb(T) is the primary signal s_λa(T) or s_λb(T) So that s "_λa(T) or s "_λbOptimize (t). FIG. 5b Output c of internal processor 32_λ(T) is the measurement signal S_λa(T) or S_λb(T ) Is subtracted from n "_λa(T) = s_λa(T) + n_λa(T) -c_λa(T) Or n "_λb= S_λb(T) + n_λb(T) -c_λbThe output signal defined as (t) is can get. The internal processor is n "_λa(T) or n "_λb(T) is primary Signal n_λa(T) or n_λbN ″ such that it is approximately equal to (t)_λa(T) or n "_λb Optimize (t).   One algorithm that can be used to adjust the transfer function of internal processor 32 is: Published by Prentice Hall and copyrighted in 1985 Bernard Widrow and Samuel Stearns muel Stearns), Adaptive Signal Processing, Chapters 6 and 12. This is the least-squares algorithm used. This book, including Chapters 6 and 12, Incorporate the entire contents as part of the text.   The adaptive processor 30 of FIGS. 5a and 5b provides a side lobe canceling of the antenna. , Pattern recognition, cancellation of common periodic interference, and long-distance telephone transmission lines. It has been successfully applied to many problems, including canceling echoes in However , N_λa(T), n_λb(T), s_λa(T) and s_λb(T) is the composite signal for measurement. No. S_λa(T) and S_λb(T) cannot be easily separated from the appropriate reference signal n Significant work is often required to find '(t) or s' (t). Real Secondary part n_λa(T) or n_λb(T) or primary signal part s_λa(T) young Or s_λbIf any of (t) can be used speculatively, techniques such as correlation cancellation No art is needed.General determination of primary and secondary reference signals   The manner in which the reference signals n ′ (t) and s ′ (t) can be determined is described below. To the detector For example, when the first signal is measured at the wavelength λa, the signal S_λa(T) is obtained Is:     S_λa(T) = s_λa(T) + n_λa(T) (1) Where s_λa(T) is the primary signal part, n_λa(T) is a secondary signal part.   The same measurement is performed simultaneously or almost simultaneously at different wavelengths λb. You get:     S_λb(T) = s_λb(T) + n_λb(T) (2) Note that the measurement signal S_λa(T) and S_λbAs long as (t) is obtained almost simultaneously, the secondary signal Component n_λa(T) and n_λb(T) is correlated. Because any random or This is because the incidental function affects each measurement value in a substantially similar manner. Substantive Primary signal component s_λa(T) and s_λb(T) also correlates with each other.   To obtain the reference signals n '(t) and s' (t), the measurement signal S_λa(T) and S_λb( t) is transformed to remove the primary signal component or the secondary signal component, respectively. According to the invention, one way to do this is to use the primary signal part s_λa(T) and s_λb(T) and the secondary signal part n_λa(T) and n_λbProportionality constant r during (t)_aas well as r_vAnd model the signal as follows:     s_λa(T) = r_as_λb(T)     n_λa(T) = r_vn_λb(T) (3) According to the signal model of the present invention, these proportionalities are related to absorption measurements and physiological measurements. Is satisfied in many measurements, including but not limited to Further In addition, according to the signal model of the present invention, for most measurements, the proportionality constant r_aAnd r_vTo It can be determined as follows:     n_λa(T) ≠ r_an_λb(T)     s_λa(T) ≠ r_as_λb(T) (4)   In equation (2), r_aAnd then subtracting equation (2) from equation (1) gives the primary signal term s_λa (T) and s_λb(T) is eliminated to yield the following single expression:     n '(t) = S_λa(T) -r_aS_λb(T)            = N_λa(T) -r_an_λb(T) (5a) This is the secondary signal part n_λa(T) and n_λb(T) is correlated with the adaptive noise Be used as a secondary criterion n '(t) in a correlation canceller such as a canceller Is a non-zero signal.   In equation (2), r_vAnd then subtracting equation (2) from equation (1) gives the primary signal term s_λa (T) and s_λb(T) is eliminated to yield the following single expression:     s' (t) = S_λa(T) -r_vS_λb(T)            = S_λa(T) -r_vs_λb(T) (5b) This is the primary signal part s_λa(T) and s_λb(T) and adaptive noise cancellation Can be used as the primary criterion s' (t) in correlation cancellers such as Serra Non-zero signal.Example of measuring primary and secondary reference signals in an absorption system   Correlation cancellation is particularly useful for many measurements collectively referred to as absorption measurements. Departure Adaptive noise based on criteria n '(t) or s' (t) determined by the Of absorption type monitor that can advantageously use correlation cancellation such as noise cancellation Determines the concentration of the energy absorbing component in the absorbing material if the absorbing material changes. Monitor. Such a change is a force or primary force that requires information. Or by random or incidental secondary forces such as mechanical forces in the material Can occur. Due to random or incidental interference such as movement, Secondary components occur. These secondary components can be determined by appropriate secondary criteria n ′ (t) or one If the next criterion s ′ (t) is known, it can be removed or derived by the correlation canceller. Wear.       A₁, A_Two, A_Three... A_NHas N different absorbing components labeled A schematic N-component absorbent material including a container 42 is shown in FIG. 6a. In FIG. 6a A₁~ A_NAre arranged in the container 42 in a substantially regular layered manner. Specific One example of an Ip absorption system is that light energy passes through container 42 and It is a system that is absorbed according to Beer Lambert's law of light absorption. wavelength For light at λa, this attenuation can be approximated by: First, take the natural logarithm of both sides, convert the signal by manipulating the terms, The signal is converted such that the signal components are combined by addition rather than multiplication. Where I₀Is the intensity of incident light energy; I is the intensity of transmitted light energy;_{i, λa}Is a wave The absorption coefficient of the i-th component in the length λa; x_i(T) is the light path of the i-th layer. Length, ie, the thickness of the material of the i-th layer through which the light energy passes; c_i(T) Thickness x_iThe concentration of the i-th component in the volume associated with (t). Absorption clerk Number ε₁~ Ε_NIs a constant known value at each wavelength. Most concentrations c₁(T) ~ c_N(T) is the optical path length x of each layer._iLike most of (t), I don't know It is typical. The total length of the light path is the individual light path length x of each layer_i(T) Is the sum of each.   If the material is not subject to any force that changes the layer thickness, the optical path length of each layer will be approximately It is constant. As a result, the light energy attenuates almost constant, so that the measurement signal Has a substantially constant offset. Typically, knowledge of the forces that perturb the material Is usually of little interest in this offset portion of the signal. Strange Includes certain unwanted signal parts obtained from almost constant component absorption when not subject to quantization , Any signal portions outside the known bandwidth of interest are removed. This is the traditional band pass It is easily achieved by over-filtering techniques. However, the force on the material Then, each component layer is affected differently from the other layers by perturbation. each Optical path length x of layer_iSome perturbation of (t) indicates desired or primary information Displacement can occur in the measurement signal. Light path length x of each layer_i(T) other Movement causes unwanted or secondary displacement that masks primary information in the measurement signal . Secondary signal components related to secondary displacement are removed to obtain primary information from the measured signal It must be. Similarly, the secondary signal component generated by the secondary displacement is calculated. If possible, the primary signal generation from the measured signal via simple subtraction or correlation cancellation techniques You can get minutes.   Correlation cancelers perturb or change the absorbing material to produce primary signal components or secondary signal components. Perturbing or changing the material differently from the force that caused any of the signal components Absorbs either the secondary signal component or the primary signal component generated by the force The material is removed from the composite signal measured after being transmitted or reflected. Explanation eyes The primary signal s_λaThe portion of the measurement signal that is considered to be (t) is the target component. Minute A_FiveAnd the damping term ε associated with_Fivec_Fivex_Five(T) and component A_FiveLayer of other component A₁~ A_Four And A₆~ A_NSuppose that each layer is affected by a different perturbation. Such a state An example of a situation is when the power that the relevant information appears to be primary is layer A_FiveAffected In addition, this is the case where the entire material receives a force that affects each layer. In this case, the ingredients A_FiveThe total force affecting each layer is the total force affecting each of the other layers. Component A, unlike the power of_FiveInformation about the layer forces and the resulting perturbations is primary Component A₁~ A_FourAnd A₆~ A_NThe attenuation term due to the secondary signal part n_{λ a} (T) is constituted. The additional force affecting the whole material is A_FiveEach layer, including layers Cause the same perturbation in_FiveThe total force in the layer of₁ ~ A_FourAnd A₆~ A_NWill have a different total perturbation than each of the other layers You.   The total perturbation affecting the layers associated with the secondary signal components is random Can often be caused by incidental forces. This allows the thickness of the layer and the light path of each layer Road length x_i(T) will be changed concomitantly, which results in a random Or secondary signal component N_λa(T) is generated. However, two Next signal part n_λaRegardless of whether (t) is adjoint or not, the secondary signal component n_λa( t) is the component A_FiveIs a component A_FiveDifferent from perturbation in layers As long as the secondary criterion n '(t) or the primary criterion determined by the processor of the invention A correlation canceller such as an adaptive noise canceller having s' (t) as one input. Can be removed or derived. The correlation canceller is the primary signal s_λa (T) or secondary signal n_λaObtain a valid approximation to any of (t). one For some physiological measurements, the primary signal The thickness associated with the component (in this example, x_Five(T)) is known or determined The concentration c of the target component_FiveDetermine (t) in most cases be able to.   The correlation canceller is used to measure two measurement signals S measured almost simultaneously._λa(T) and S_λb Use the secondary criterion n '(t) or primary s' (t) determined from (t). S_λa (T) is determined as described above in equation (7). S_λb(T) is a different wave It is determined similarly for the length λb. Either the secondary criterion n '(t) or the primary criterion s' (t) To find out, the attenuated transmitted energy is measured at two different wavelengths λa and λ measured at b and transformed via log transformation. Signal S_λa(T) and S_λb (T) can be written as follows (log conversion possible):   Further transformation of the signal is performed as in equation (3)._aAnd r_vThe signal mode of the present invention that defines This is a proportional relation by Dell, whereby the noise criterion n ′ (t) and the primary criterion s ′ (t) are Can decide. These are as follows: ε_{Five, λa}= R_aε_{Five, λb}                                              (12a) n_λa= R_vn_λb                                                  (12b) here, n_λa≠ r_an_λb                                                  (13a) ε_{Five, λa}≠ r_vε_{Five, λb}                                              (13b) It is. Often, both equations (12) and (13) are satisfied simultaneously. In equation (11), r_aMultiply by And subtracting the result from equation (9) yields a zero, which is the linear sum of the secondary signal components. No secondary criteria are obtained.   In equation (11), r_v, And subtracting the result from equation (9) gives the primary signal component A primary criterion that is a linear sum is obtained. s' (t) = S_λa(T) -r_vS_λb(T)        = S_λa(T) -r_vs_λb(T) (14b)        = C_Fivex_Five(T) ε_{Five, λa}-R_vc_Fivex_Five(T) ε_{Five, λb}             (15b)        = C_Fivex_Five(T) [ε_{Five, λa}-R_vε_{Five, λb}] (16b)   A sample of either the secondary reference n '(t) or the primary reference s' (t) and the measurement signal S_{λ a} (T) or S_λbAny sample of (t) is shown in FIGS. 5a and 5b. An example is input to a correlation canceller 27 such as an adaptive noise canceller 30 shown in FIG. A preferred example is the preferred correlation canceller using the implementation of the joint process estimator ( Preferred Correlation Canceler Using a Joint Process Estimator Implement ation) "herein. Correlation canceller 27 Secondary part n of the measurement signal_λa(T) or n_λb(T) or primary component s_λa(T) Or s_λb(T) is removed and a valid approximation s "of the primary signal is obtained._λa(T) ≒ ε_{Five λa}c_Fivex_Five(T) or s "_λb(T) ≒ ε_{Five λb}c_Fivex_Five(T) or two A valid approximation n ”of the next signal_λa(T) ≒ n_λa(T) or n "_λb(T) ≒ n_λb (T) is obtained. If a primary signal is available, the near Similar_λa(T) or s "_λbFrom (t) to concentration c_Five(T) is determined as follows Can be: c_Five(T) $ s "_λa(T) / ε_{5, λa}x_Five(T) (17a) c_Five(T) $ s "_λb(T) / ε_{Five, λb}x_Five(T) (17b) As described above, the absorption coefficient is constant at each wavelength λa and λb, and in this example, Thickness x related to primary signal component_Five(T) is often known or time Component A since it can be determined as a function between_FiveConcentration c_FiveCalculate (t) Can beDetermination of concentration or saturation in a volume containing more than one component   Referring to FIG. 6b, another material having N different components arranged in multiple layers is shown. Is done. In this material, two components A_FiveAnd A₆Is the thickness x_5,6(T) = x_Five(T) + x₆(T) is found in one layer having substantially randomly located in this layer ing. This corresponds to the component A in FIG._FiveAnd A₆Similar to the joining of layers. Component A_Fiveas well as A₆Bonding between layers such as bonding of layers is suitable when the overall force is the same, In this case, the optical path length x of the two layers_Five(T) and X₆(T) has the same change Will be lost.   Often, a component within a given thickness that contains more than one component and is subject to unique forces It is desirable to find the concentration or saturation of the minute, ie the percent concentration. Given boli The measurement of the concentration or saturation of a component in a volume Number components are perturbed or changed in the same way by receiving the same total force Can be done if One component in one volume containing many components To determine the degree of saturation, the same number of measurement signals as the component absorbing the incident light energy is required. It is important. Components that do not absorb light energy are not important in determining saturation. Will be understood. To determine the concentration, the same as the component that absorbs the incident light energy A number signal and information about the sum of the concentrations are required.   Often the thickness under unique operation contains only two components. For example, A_Five And A₆A in a given volume containing_FiveTo know the concentration or saturation of Sometimes it is better. In this case, the primary signal s_λa(T) and s_λb(T) is A_Five And A₆A in the volume, including terms related to_FiveOr A₆Concentration or saturation Can be measured. Here, the measurement of the degree of saturation will be described. A_FiveAnd A₆Both A in the containing volume_FiveThe concentration of A_Five+ A₆= 1, ie the selected specific measurement No component in the volume that does not absorb incident light energy at a constant wavelength The case can also be determined. Measurement signal S_λa(T) and S_λb(T) is It can be written as follows (log conversion possible): S_λa(T) = ε_{Five, λa}c_Fivex_5,6(T)              + Ε_{6, λa}c₆x_5,6(T) + n_λa(T) (18a)              = S_λa(T) + n_λa(T) (18b) S_λb(T) = ε_{Five, λb}c_Fivex_5,6(T)              + Ε_{6, λb}c₆x_5,6(T) + n_λb(T) (19a)              = S_λb(T) + n_λb(T) (19b)   As shown in FIG. 6c, each contains the same two components but undergoes another operation. It is often the case that more than one thickness is present in one medium. For example , A_FiveAnd A₆A in a given volume containing_FiveTogether with the concentration or saturation of A_FiveAnd A₆A having the same components as_ThreeAnd A_FourA in a given volume containing_Threeof It may be desirable to know the concentration or saturation. Again, the primary signal s_{λ a} (T) and s_λb(T) is A_FiveAnd A₆And the secondary signal n_λa (T) and n_λb(T) is A_ThreeAnd A_FourIncluding terms related to both. A_Three And A_FourLayers do not enter the primary signal equation, because A_ThreeAnd A_FourIs a different lap Random or concomitant, not correlated with wave number or with the force associated with the primary signal This is because it is considered to be perturbed by a typical secondary force. Ingredients 3 and 5 and Since minutes 4 and 6 are assumed to be the same, they have the same absorption coefficient (ie, ε_{Three λa}= Ε_{Five λa}; Ε_{Three λb}= Ε_{Five λb}; Ε_{Four λa}= Ε_{6 λa}; Ε_{Four λb}= Ε_{6 λb}). However But generally speaking, A_ThreeAnd A_FourIs A_FiveAnd A₆Has a different concentration from Will have a certain degree of saturation. Therefore, a single component in one medium It may have one or more associated degrees of saturation. Primary and secondary signals according to this model Can be written as: s_λa(T) = [ε_{Five, λa}c_Five+ Ε_{6, λa}c₆] X_5,6(T) (20a) n_λa(T) = [ε_{Five, λa}c_Three+ Ε_{6, λa}c_Four] X_3,4(T) + n_λa(T) (20c) s_λb(T) = [ε_{Five, λb}c_Five+ Ε_{6, λb}c₆] X_5,6(T) (21a) n_λb (t) = [ε_{Five, λb}c_Three+ Ε_{6, λb}c_Four] X_3,4(T)                 + N_λb(T) (21c) Signal n_λa(T) and n_λb(T) is the secondary signal n except for omitting the layers 3 and 4._λa (T) and n_λbSame as (t).   Constant undesired second-order signal part obtained from nearly constant absorption of components when not perturbed Signals outside the known bandwidth of interest, including those associated with the primary or secondary signal Approximation to either the primary or secondary signal within the bandwidth of interest Should be removed to determine This is a traditional bandpass filter It is easily achieved with the help of a technic. As in the previous example, the Total perturbations or changes affecting a given layer are random or adjoint The thickness of each layer or the optical path length x of each layer_i(T) is incidental And a random or concomitant secondary signal component n_λa(T) is generated You. Secondary signal part n_λaComponent A, whether (t) is concomitant or not_FiveAnd A₆ Is a component A_FiveAnd A₆Limit different from perturbation in the layer of And the secondary criterion n '(t) or the primary criterion s' (t) determined by the processor of the present invention. Through a correlation canceller such as an adaptive noise canceller that takes No. component n_λa(T) can be removed or derived. Correlation canceller , An optional secondary signal component n_λa(T) and n_λb(T) or primary component s_λa(T) And s_λbAdvantageously, either (t) is derived from equations (18) and (19) or from equations (20) and (21) Can be removed. Also in this case, the correlation canceller uses the primary reference s ′ (t) or Any sample of the secondary criterion n '(t) and the composite signal S of equations (18) and (19)_λa(T ) Or S_λbAny sample of (t) is required.Determination of primary and secondary reference signals for saturation measurement   According to one aspect of the invention, the measurement signal S_λa(T) and S_λbFrom (t), the reference signal s ′ ( One way to determine t) or n '(t) is what is called the constant saturation method. . In this method, A_FiveAnd A₆A in the volume containing_FiveSaturatio n) and A_ThreeAnd A_FourA in the volume containing_ThreeIs saturated for some time Assume that it remains almost constant. That is, the degree of saturation is determined by the measurement signal S_λa(T) And S_λbIt is substantially constant for many samples in (t). Saturation generally changes relatively slowly in physiological systems, so This assumption holds in the sample of. The assumption of constant saturation is equivalent to assuming: Because the other terms in Equations (23a) and (23b) are constants, that is, 1. Because.   Using this assumption, the secondary reference signal n ′ (t) and the primary Proportional constant r allowing determination of reference signal s' (t)_aAnd r_vIs:   Where n_λa(T) ≠ r_a(T) n_λb(T) (30a)   as well as,   Where s_λa(T) ≠ r_v(T) s_λb(T) (30b) It is.   According to the present invention, in many cases both equations (26) and (30) are satisfied simultaneously. , The proportionality constant r_aAnd r_vCan be determined. Furthermore, the absorption coefficient ε at each wavelength_{Five λa}, ε_{6 λa}, Ε_{Five λb}, Ε_{6 λb}Is constant, and the central assumption of the constant saturation method is c_Five (T) / c₆(T) and c_Three(T) / c_Four(T) over many sample periods That is, it is constant. Therefore, the primary signal output from the correlation canceller Or a new approximation to either the secondary signal New proportional constant r_aAnd r_vCan be determined. Therefore, the processor of the present invention measures Constant signal S_λa(T) and S_λbWith respect to the sample set substantially immediately before (t) The primary signal s found by the correlation canceller_λa(T) and s_λb(T) Be young Is the secondary signal n_λa(T) and n_{λb (}An approximation to any of t) is used, Measurement signal S_λa(T) and S_λbThe proportionality constant r for the next sample set of (t)_a And r_vIs calculated.   In equation (19), r_aBy subtracting the resulting equation from equation (18) No secondary reference signal is obtained.   n '(t) = S_λa(T) -r_aS_λb(T)          = N_λa(T) -r_an_λb(T) (31a)   In equation (19), r_vBy subtracting the resulting equation from equation (18) No primary reference signal is obtained.     s' (t) = S_λa(T) -r_vS_λb(T)            = S_λa(T) -r_vs_λb(T) (31b)   Using the constant saturation method for patient monitoring, the initial proportionality factor is Can be determined as described below. Patients must remain inactive even during the initialization period. No need. Proportional coefficient r_aAnd r_vOnce the value of is determined, the correlation canceller It can be used with n '(t) or the primary criterion s' (t).Determination of signal coefficients of primary and secondary reference signals using constant saturation method   According to one aspect of the present invention, as shown in FIG._{λ b} (T) = s_λb(T) + n_λb(T) has a plurality of signal coefficients r₁, R_Two... r_nEach of Each of the obtained equations is multiplied by the first measurement signal S_λa(T) = s_λa(T) + n_λa(T) By subtracting r = r₁, R_Two... r_nGet multiple reference signals for As such, the reference processor 26 of FIGS. 4a and 4b of the present invention can be configured. You.   R '(r, t) = s_λa(T) -rs_λb(T) + n_λa(T) -rn_λb(T)                                                                   (32) In other words, multiple signal coefficients are used to indicate a cross section of possible signal coefficients. Selected.   The primary reference s ′ (t) or the secondary reference n ′ (t) is obtained from the plurality of reference signals of the above equation (32). To determine either, a plurality of assumed signal coefficients r₁, R_Two... r_nFrom the signal coefficient r_aAnd r_vIs determined. Primary signal part s_λa(T) and s_λb(T) or secondary credit No. part n_λa(T) and n_λb(T) is, for example, a reference function R ′ (r, t) as follows: Any of them will be canceled or almost canceled if assigned to So that the coefficient r_aAnd r_vIs selected.     s_λa(T) = r_as_λb(T) (33a)     n_λa(T) = r_vn_λb(T) (33b)     n ′ (t) = R ′ (r_a, T) = n_λa(T) -r_an_λb(T) (33c)     s '(t) = R' (r_v, T) = s_λa(T) -r_vs_λb(T) (33d)   In other words, the coefficient r_aAnd r_vIs the maximum correlation between the primary and secondary signal parts. Selected with a value that reflects the small value. In practice, the measurement signal S_λa(T) and S_λb(T ) Primary signal part s_λa(T) and s_λb(T) or secondary signal part n_λa(T ) And n_λbThere is usually not enough information about (t) beforehand. This Due to lack of information of₁, R_Two... r_nWhich is the signal coefficient r_a= S_λa( t) / s_λb(T) and r_v= N_λa(T) / n_λbCorresponding to (t) Is difficult to determine.   Multiple coefficients r₁, R_Two... r_nFrom the signal coefficient r_aAnd r_vOne way to determine is , As shown in FIG._λa(T) or S_λbOne of (t) A corresponding first input is received and a plurality of reference signals R ′ (r₁, T), R ′ (r_Two, T), .. . R '(r_n, T), the adaptive noise canceller receiving a second input corresponding to each one successively. A correlation canceller 27 such as a canceller is used. Reference signal R ′ (r₁, T), R ′ (r_Two , T), ... R '(r_n, T), the corresponding output of the correlation canceller 27 Is input to a “square” operation 28 which squares the output of the correlation canceller 27. It is. The output of the square operation 28 is provided to an integrator 29 and An output signal (sum of squared values) is formed. The accumulated output signal is then output to the extreme value detection means. 31 is input. The purpose of the extreme value detecting means 31 is as shown in FIGS. 7B and 7C. By observing the coefficient that gives the maximum value in the output signal, r₁, R_Two... r_nof Set to signal coefficient r_aAnd r_vIs to choose. In other words, the correlation scan The coefficient that provides the maximum integrated output such as energy or power from the serra 27 is The minimum correlation between the primary and secondary signal parts according to the signal model of the present invention is Signal coefficient r_aAnd r_vCorresponding to Minimum value or minimum value in the accumulated output signal Providing change₁, R_Two... r_nThe coefficients from the set of_aAnd r_vToss System geometries that need to be positioned You.   Using a plurality of coefficients in the processor of the present invention together with the correlation canceller 27 Signal coefficient r_aAnd r_vIs determined by using the characteristics of the correlation canceller. Can be demonstrated. x, y, and z are arbitrary sets of three signals that vary with time. Given that, the properties (Property) C (x, y) of some correlation cancellers are It can be defined as:   Property (1) C (x, y) = 0 When x, y are correlated (34a)   Property (2) C (x, y) = xx When x and y are not correlated (34b)   Property (3) C (x + y, z) = c (x, z) + c (y, z) (34c) If the properties (1), (2), and (3) are used, the measurement signal S_λa(T) or S_{λ b} (T) and a plurality of reference signals R ′ (r) corresponding to one of the inputs.₁, T), R '( r_Two, T), ... R '(r_n, T) having a second input corresponding to each successive one of The energy or power output of the canceller is the primary criterion s' (t) and the secondary criterion Signal coefficient r required to generate n '(t)_aAnd r_vDemonstrating that you can determine It is easy. The measurement signal S as a first input to the correlation canceller_λa(T), A plurality of reference signals R ′ (r₁, T), R ′ (r_Two, T), ... R '(r_n, T) , The output C (S) of the correlation canceller for j = 1, 2,._λa( t), R '(r_j, T)) can be written as: C (s_λa(T) + n_λa(T), s_λa(T) -r_js_λb(T) + n_λa(T)- r_jn_λb(T)) (35) Here, j = 1, 2,..., N, and the following equation was used. R '(r, t) = S_λa(T) -rS_λb(T) (36) S_λa(T) = s_λa(T) + n_λa(T) (37a) S_λb(T) = s_λb(T) + n_λb(T) (37b)   By using Property (3), we can expand Equation (35) into two terms You. C (S_λa(T), R '(r, t)) = C (s_λa(T), s_λa(T) -rs_λb (T) + n_λa(T) -rn_λb(T)) + C (n_λa(T), s_λa(T) -rs_λb (T) + n_λa(T) -rn_λb(T)) (38) Using Property (1) and (2), the output of the correlation canceller is defined as follows: It is. C (S_λa(T), R '(r_j, T)) = s_λa(T) δ (r_j-R_a) + N_λa(T) δ (r_j-R_v) (39) Here, δ (x) is a unit impulse function.         When δ (x) = 0 x ≠ 0         δ (x) = 1 When x = 0 (40)   Output C (S_λa(T), R '(r_j, T)) is the time variable t Can be omitted by calculating the energy or power of Correlation The output energy of the canceller is defined as follows. E_λa(R_j) = ∫C^Two(S_λa(T), R '(r_j, T) dt                       = Δ (r_j-R_a) ∫s^Two _λa(T) dt                         + Δ (r_j-R_v) ∫n^Two _λa(T) dt (41a)   The measurement signal S as a first input to the correlation canceller_λbSelect (t) and likewise A plurality of reference signals R ′ (r₁, T), R ′ (r_Two, T), ... R '(r_n, T) It should be understood that the selection was equally successful. In this case, the correlation canceller energy Energy output is E_λb(R_j) = ∫C^Two(S_λb(T), R '(r_j, T) dt                     = Δ (r_j-R_a) ∫s^Two _λb(T) dt                     + Δ (r_j-R_v) ∫n^Two _λb(T) dt (41b) It is.   In actual situations, not only continuous time measurement signals but also discrete time measurement signals It should also be understood that it can be used. A system for performing a discrete transform according to the invention The system will be described with reference to FIGS. When using discrete time measurement signals , Trapezoidal rule, midpoint rule, Tick's law, Simpson's Use an integral approximation method such as approximation or other techniques to reduce the energy Alternatively, it is possible to calculate the power output. In the case of discrete time signal measurement The energy output of the correlation canceller should be written using the trapezoidal rule as Can: t_iIs the ith discrete time, t₀Is the initial time, t_nIs the final time, Δt is the discrete time Is the time between the measured sample and the sample.   The energy function of FIG. 7b defined above corresponds to the measurement signal S_λa(T) or S_λb (T) and a plurality of reference signals R ′ (r₁, T), R ′ (r_Two, T), ... R '(r_n, T) many Indicates that the output of the correlation canceller is normally zero. However, the energy function is based on the reference signal R '(r_j, T) the primary signal part s_λa(T) and s_λb(T) or secondary signal part n_λa(T) and n_λb(T) R corresponding to any cancellation_jIs not zero for the value of. These values are Number r_aAnd r_vCorresponding to   Primary signal part s_λa(T) and s_λb(T) or secondary signal part n_λa(T) and n_λbThere is a time when any of (t) is equal to or almost zero I want to be understood. In such a case, only one signal coefficient value is Provide maximum energy or power output. The biggest energy of correlation canceller Ambiguous because there can be more than one signal coefficient value that provides power or power output. Situations can occur. The signal coefficient together with the criterion function R ′ (r, t) is a primary Providing any of the following criteria cannot be immediately clarified. In this case Requires the use of physical system constraints. For example, a pulse oximeter Measurements show that arterial blood with the characteristics of primary plethysmographic waves is Alternatively, it has a greater oxygen saturation than venous blood having the characteristics of a random signal. Therefore, in pulse oximetry, the primary signal ratio r due to arterial pulsation_a= S_λa(T) / s_λb (T) is the smaller of the two signal coefficient values, while mainly the venous blood Secondary signal ratio r due to liquid flow_v= N_λa(T) / n_λb(T) is two signals This is the larger of the coefficient values. λa = 660 nm and λb = 910 nm Suppose there is. In the actual implementation of multiple reference signals and cross-correlation techniques, the Property (Characteristics) The ideal features listed above as (1), (2) and (3) are exactly Not satisfied, an approximation of that. Thus, this embodiment of the invention is actually implemented And the energy curve of the correlation canceller shown in FIG. Rather than consist of a function, a finite width can be associated as shown in FIG. 7c.   Signal coefficient that produces the maximum energy or power output from the correlation canceller It should also be understood that there can be more than two values. This situation , Occurs when each measurement signal contains more than two components, where each component is Related by ratio: here, f_{λa, i}(T) = r_if_{λb, i}(T)                                 i = 1 ...., nr_i≠ r_j It is.   Therefore, use of a reference signal together with a correlation canceller such as an adaptive noise canceller Decomposes the signal into two or more signal components, each related by one ratio can do.Preferred correlation canceller using implementation of joint process estimator   Either the secondary criterion n '(t) or the primary criterion s' (t) Once the decision is made, the correlation canceller can be used in either hardware or software. This can be done in any case. The preferred implementation of the correlation canceller is a joint process estimator. Performing an adaptive noise canceller using stages.   The internal processor 32 described above with the adaptive noise canceller of FIGS. It is relatively easy to perform the least mean square (LMS) of Lack of desirable application speed for most applications of dynamic monitoring You. Therefore, in one embodiment, the least squares lattice type combined process estimator model Use a faster adaptive noise cancellation method called. Joint process estimation Means 60 is illustrated in FIG. 8 and Prentice-Hall is copyrighted "Adaptive F" by Simon Haykin, published in 1986 See Chapter 9 of "ilter Theory". Support the entire contents of this book, including Chapter 9 As part of the present invention.   The function of the combining process estimator is that the measurement signal S_λa(T) or S_λb(T) Secondary signal part n_λa(T) or n_λb(T) or primary signal part s_λa(T ) Or s_λb(T), and the primary signal approximation s "_λa(T) Be young Is s "_λb(T). Therefore, the joint process estimator is , Primary signal s_λa(T) or s_λb(T) or secondary signal n_λa(T) Be young Is n_λbEstimate any value of (t). Input to the joint process estimating means 60 Is either the secondary criterion n ′ (t) or the primary criterion s ′ (t) and the composite measurement signal S_{λ a} (T) or S_λb(T). Its output is the secondary or primary signal The signal S from which either has been removed_λa(T) or S_λbValid approximation to (t) Ie, s_λa(T), s_λb(T), n_λa(T), n_λbValid for (t) It is an approximation.   The combination process estimating means 60 shown in FIG. Used together with the data 80. Either the secondary reference signal n '(t) or the primary reference s' (t) Are input to the least-squares lattice type prediction means 70, and the measurement signal S_λa(T) or S_λb(T) is input to the regression filter 80. For convenience in the following description First, the primary part s_λa(T) or secondary part n_λaThe measurement signal for which one of (t) is estimated is S_λa(T). Only Then S_λb(T) is input to the regression filter 80 to obtain the primary part s of this signal._λb(T ) Or secondary part n_λb(T) can also be estimated.   The combining process estimator 60 determines whether the reference n ′ (t) or s ′ (t) and the measurement signal S_λa Remove all frequencies present in both of (t). Secondary signal part n_λa(T) is Usually the primary signal part s_λaIncludes frequencies unrelated to the frequency of (t). Second letter No. part n_λa(T) is the primary signal part s_λaIf it contains exactly the same spectrum as (t) It cannot be said. However, s_λa(T) and n_λa(T) is the same In situations (likely unlikely) involving similar spectra, this method will give accurate results. I can't. Functionally, the combining process estimator 60 comprises a second signal part n_λa(T ) Or primary signal part s_λaReference input signal n '(t) correlated with any of (t) Or s' (t) and the input signal S_λaComparing (t), all frequencies that are the same Remove. Therefore, the joint process estimating means 60 is a dynamic multi-notch filter. And secondary signal components n as they change concomitantly with patient movement_λa(T) Primary signal component when removing frequencies in the signal or changing with the patient's arterial pulsation s_λaFor example, the frequency in (t) is removed. Thereby, the primary signal s_λa(T ) Or secondary signal n_λaSignal having substantially the same spectral components as any of (t) No. is obtained. Therefore, the output s ″ of the combining process estimation means 60_λa(T) Be young Is n "_λa(T) is the primary signal s_λa(T) or secondary signal n_λa(T) no This is a very useful approximation to that.   The combining process estimator 60 starts at the zero stage as shown in FIG. The stage ends with the m-th stage. Each stage except for the zero stage The tage is the same as all other stages. Zero stage is a joint process estimator Stage 60 is an input stage. The first stage to the m-th stage are By acting on the signal generated at the (m-1) th stage, Effective primary signal approximation s "_λa(T) or secondary signal approximation n ″_λa(T) is the m-th Generated as stage output.   The least-squares lattice type prediction means 70 includes registers 90 and 92 and an addition element 1 00 and 102 and a delay element 110. Registers 90 and 92 are Reflection coefficient Γ_{f, m}(T) and the back reflection coefficient Γ_{b, m}(T) Is the reference signal n '(t) or s' (t) and these reference signals n '(t) or s' (t) Is multiplied by the signal obtained from Each state of the least squares grid type prediction means Is the forward prediction error f_m(T) and backward prediction error b_m(T) is output. The subscript m is Show the stage.   For each sample set, ie the measurement signal S_λaAbbreviated as one sample of (t) For one sample of the simultaneously derived reference signal n '(t) or s' (t), The sample of the reference signal n ′ (t) or s ′ (t) is used as the least square lattice type prediction means 70. Is input to Zero stage forward prediction error f₀And zero-stage backward prediction error b₀(t) is set equal to the reference signal n '(t) or s' (t). rear Prediction error b₀(t) is delayed in the first stage of the least square lattice type prediction means 70. Delayed by the delay element 110 by one sample period. Therefore, the criterion n '(t) Alternatively, the value immediately before s ′ (t) is calculated using the first stage delay element 110. Used in The forward prediction error of the zero stage is the forward reflection coefficient value Γ_{f, 1}( t) Delayed zero stage backward prediction error b multiplied by register 90 value₀(t-1) Is added to the negative value, the forward prediction error f of the first stage₁(t) is generated. further , Zero stage forward prediction error f₀(t) is the back reflection coefficient Γ_{b, 1}(T) Register 9 Multiplied by 2 and the delayed backward prediction error b of the zero stage₀added to (t-1) , The backward prediction error b of the first stage₁(T) is generated. Least-squares lattice predictor At each stage m after stage 70, the previous forward prediction error value f_m-1(t) and 1 Backward prediction error value b delayed by the sample period_m-1Using (t-1), the current status Forward prediction error f_m(T) and backward prediction error b_mThe value of (t) is generated.   Backward prediction error b_m(T) is supplied to the cooperating stage m of the regression filter 80. Where the regression coefficient value κ_{m, λa}(T) is input to the register 96. For example, times In the zero stage of the recursive filter 80, the zero stage backward prediction error b₀Zero at (t) Stage regression coefficient k_{0, λa}The value of the register 96 shown in FIG. The signal S_λaSubtract from the measured value of (t) to estimate the first stage Error signal e_{1, λa}(T) is generated. First stage estimation error signal e_{1, λa}(T) Is the first approximation to either the primary signal or the secondary signal. This first Estimated stage error signal e_{1, λa}Is input to the first stage of the regression filter 80. You. First stage regression coefficient κ_{1, λa}The first multiplied by the value of the register 96 of (t) Stage backward prediction error b₁(T) is the first stage error signal e_{1, λa}Subtract from (t) And the estimation error e of the second stage_{2, λa}(T) is generated. Second stage promotion Constant error signal e_{2, λa}(T) is the primary signal s_λa(T) or secondary signal n_λa(T ) Is a somewhat better second approximation to either of these.   Primary signal s_λa(T) or secondary signal n_λaA valid approximation e to (t)_{m λa} Until (t) is determined, the least squares lattice type prediction means 70 and the regression filter 8 At 0, the same process is repeated for each stage. Forward prediction error f_m(T) , Backward prediction error b_m(T), estimated error signal e_{m, λa}Each of the above signals, including (t), Forward reflection coefficient at each stage m_{f, m}(T), back reflection coefficient Γ_{b, m}(T), times Regression coefficient κ_{m, λa}It is necessary to calculate the values of the registers 90, 92 and 96 of (t). is there. Forward reflection coefficient Γ_{f, m}(T), back reflection coefficient Γ_{b, m}(T) and regression coefficient κ_{m, λ a} To calculate the values of the registers 90, 92 and 96 of (t), the forward prediction error f_m( t), backward prediction error b_m(T), estimated error signal e_{m λa}(T) In addition to the signal A large number of intermediate variables (not shown in FIG. 8) based on the values shown in FIG. You.   The intermediate variable is the weighted sum of the squares of the forward prediction error F_m(T), the backward prediction error Squared weighted sum B_m(T), scalar parameter Δ_m(T), conversion factor γ_m (T) and another scalar parameter ρ_{m, λa}(T). Weight of forward prediction error Finding total F_m(T) is defined as follows: In the above formula, λ having no wavelength identifier, ie, a or b, is a constant power regardless of wavelength. It is a calculated value, typically 1 or less, that is, λ ≦ 1. Backward prediction error B_m(T) The weighting error is defined as: In the above formula, λ having no wavelength identifier, ie, a or b, is a constant power regardless of wavelength. It is a calculated value, typically 1 or less, that is, λ ≦ 1. Haykin mentioned above As described in Section 9.3 of the book and later defined in equations (59) and (60). So that these weighted sum intermediate error signals can be determined more easily can do.Explanation of the joining process estimation means   The operation of the combining process estimating means 60 is as follows. Binding process When the estimating means 60 is turned on, the parameter Δm_-1(T), the forward prediction error signal Weighted sum F_m-1(T), weighted sum B of backward prediction error signals_m-1(T), para Meter ρ_{m, λa}(T) and the estimation error e of the zero stage_{0, λa}Intermediate variables including (t) Initial values and numbers are initialized to zero, some to zero, Some are initialized to the number δ of Δ_m-1(0) = 0; (46) F_m-1(0) = δ; (47) B_m-1(0) = δ; (48) ρ_{m, λa}(0) = 0 (49) e_{0, λa}(T) = S_λa(T) When t ≧ 0 (50)   As shown in FIG. 8, after initialization, the measurement signal S_λa(T) or S_λb(T ) And either the secondary criterion n ′ (t) or the primary criterion s ′ (t) It is input to the combining process estimating means 60. Forward prediction error for zero stage Signal f₀(t) and backward prediction error signal b₀(t) and the weighted sum F of the forward error signal₀( t) and the weighted sum B of the backward error signal₀Intermediate variables including (t) and conversion factor γ₀( t) is calculated by: f₀(T) = b₀(T) = n '(t) (51a) F₀(T) = B₀(T) = λF₀(T-1) + | n '(t) |^Two              (52a) γ₀(T-1) = 1 (53a) If the secondary criterion n '(t) is used, then: f₀(T) = b₀(T) = s' (t) (51b) F₀(T) = B₀(T) = λF₀(T-1) + | s' (t) |^Two              (52b) γ₀(T-1) = 1 (53b) Even when the primary criterion s ′ (t) is used, λ without the wavelength identifier a or b It is a constant multiplication value that is not relevant.   Forward reflection coefficient at each stage Γ_{f, m}(T), back reflection coefficient Γ_{b, m}(T) and And regression coefficient κ_{m, λa}The values of registers 90, 92 and 96 in (t) are then Set according to the output of the page. Forward reflection coefficient in the first stage Γ_{f, 1}( t), the back reflection coefficient Γ_{b, 1}(T) and regression coefficient κ_{1, λa}(T) register 90, The values of 92 and 96 use the values at the zero stage of the joint process estimator 60. And set according to the algorithm. In each stage of m ≧ 1, Data Δ_m-1(t), forward reflection coefficient Γ_{f, m}(T) the value of the register 90, the back reflection coefficient Γ_{b, m} (T) value of register 92, forward error signal f_m(T) and b_m(T), Haiki Weighted sum of the squares of the forward prediction error F treated in section 9.3 of_{m, t} (T), the square of the backward prediction error, treated in Section 9.3 of the Hikin reference. Weighted sum B_{b, m}(T), conversion factor γ_m(T), parameter ρ_{m, λa}(T) , The regression coefficient κ_{m, λa}(T) Register 96 value and estimated error e_{m + 1, λa}(T); The intermediate value and the register value including the value of are set according to: here,(^*) Indicates a complex conjugate.   From these equations, the error signal f_m(T), b_m(T), e_{m, λa}(T) is squared Are multiplied by each other, and by squaring the error, Δ_m-1New intermediate errors such as (t) A value is generated. The error signal and the intermediate error value are as shown in the above equations (54) to (64). Are recursively joined together. They minimize the error signal in the next stage Interact as you do.   Primary signal s_λa(T) or secondary signal n_λaValid for any of (t) After the approximation is determined by the combining process estimator 60, the measurement signal S_λa(T) Includes samples and samples of either the secondary criterion n '(t) or the primary criterion s' (t) The next set of samples is input to the combining process estimator 60. Forward reflection coefficient Γ_{f, m}(T) and the back reflection coefficient Γ_{b, m}(T) Registers 90 and 92 values and regression coefficient κ_{m, λa}The value of the register 96 at (t) is the value of the previously input S_λaSample of (t) Primary signal part s_λa(T) or secondary signal part n_λaEstimate any of (t) The reinitialization process occurs again, as it reflects the multiplication values needed to perform No. Therefore, using the information from the previous sample, Either the primary signal part or the secondary signal part of the reset is estimated.   In a more numerically stable and preferred embodiment of the joint process estimator described above, A normalization combined process estimator is used. This version of the combined process estimator The joint process estimator described above is used so that the normalized variable is between -1 and 1. Normalize some variables. The normalized combined process estimator uses the following conditions: By redefining the variables defined by The operation is performed as shown in the title 12.   With this modification, equations (54) to (64) can be converted into the following normalization equations. Initialization of means for estimating the normalized joint process   The reference noise input at the time index t is N (t), and the time index t is Assuming that the combined signal obtained by adding the inputted noise is defined as U (t), (See Hykin's text on page 619):       1. At time t = 0, the algorithm is initialized as follows       2. At each time t ≧ 1, various zero-order variables are generated as follows: You.       3. For regression filtering, at time index t = 0, To initialize the algorithm.       4. Generate a zero-order variable at each time t ≧ 1   Therefore, the normalized joint process estimator is used for more stable systems. Can be done.   In another embodiment, the correlation cancellation is illustrated in FIG. Simon Ha, published in 1986, copyrighted by Prentice-Hall This is described in detail in Chapter 18 of "Adaptive Filter Theory" by Simon Haykin. This is performed using a QRD algorithm.   The following equation applied from the Hikin reference corresponds to the QRD-LSL diagram of FIG. (This figure is also applied from the Hikin reference.) Calculation     a. Prediction: time t = 1, 2,..., And prediction order m = 1, 2,. Calculate the following for the final predicted order):     b. Filtering: orders m = 0, 1,..., M−1 and times t = 1, 2, The following is calculated for ...       5. Initialization     a. Initialization of auxiliary parameters: For orders m = 1, 2,. Cut.     b. Lightly constrained initialization: For orders m = 0, 1,. To     Here, δ is a small positive constant.     c. Data initialization: For t = 1, 2,..., The following is calculated.     Here, μ (t) is an input at time (t), and d (t) is time (t). Response.Flow chart of joint process estimation   This is used to determine the reference n ′ (t) or s ′ (t) input to the correlation canceller. In a signal processor such as a physiological monitor with the reference processor of the invention, the coupling processor An adaptive noise canceller of the type of process estimator 60 is a source having an iterative loop. It is generally executed via a software program. One iteration of the loop , Analogous to the single stage of the combined process estimator shown in FIG. Therefore, Lou If the loop is repeated m times, it becomes equal to m stages of the joint process estimator 60.   Measurement signal S_λa(T) primary signal part s_λa(T) or secondary signal part n_λa FIG. 9 shows a flowchart of a subroutine for estimating (t). This flow The criterion plate determines which of the secondary criterion n '(t) or primary s' (t) The function of the processor is shown. The flowchart of the combined process estimation means Executed by the software.   When the physiology monitor is turned on, block 120 indicates “Initialize Noise Canceller. , The first initialization is executed. Initialize all cash registers Stars 90, 92 and 96 and delay element variable 110 are described above in equations (46)-(50). Is set to the specified value.   Then, the composite measurement signal S_λa(T) and S_λbThe set of simultaneous samples in (t) , Are input to a subroutine shown in the flowchart of FIG. Then, delay An update of each time of the increment program variable occurs, which occurs at block 130 "[Z^-1[Update element time] operation. Delay element The value stored in each of the Is set to the value Therefore, the backward prediction error b of the zero stage₀(t) is the first Stored as a delay element variable of the first stage, the first stage backward prediction error b₁(t ) Is stored as a delay element variable in the second stage, and so on. Done.   Then, the measurement signal sample S_λa(T) and S_λbUsing the set of (t), the ratio The reference signal is obtained using a ratiometric or constant saturation method described above. this is, Block 140 "reference for two measurement signal samples [n '(t) or s' (t)] calculation.   Next, as indicated by the "zero stage update" operation of block 150, The page is updated. Zero stage backward prediction error b₀(t) and zero stage Forward prediction error f₀(t) is equal to the value of the reference signal n ′ (t) or s ′ (t). Set. Further, a weighted sum F of forward prediction errors_m(T) and backward prediction Error weighting error B_m(T) equals the value defined by equations (47) and (48) Is set as follows.   Then, as indicated by the "m = 0" operation of block 160, the loop counter m Is initialized. Used by the subroutine corresponding to the flowchart of FIG. The maximum value of m that defines the total number of stages is also defined. Typically, the primary signal Or a criterion to converge to the best approximation for any of the secondary signals. A loop is configured to stop the repetition once the data estimating means 60 is satisfied. Further Then, the maximum number of loop iterations at which the loop stops an iteration may be selected. Physiology of the present invention In a preferred embodiment of the biological monitor, the maximum number of iterations is selected from m = 6 to m = 10. Advantageously.   In the loop, the order of the m-th cell of the LSL lattice type in block 170 of FIG. Update operation, as shown by the forward reflection coefficient Γ_{f, m}(T) and back reflection Number_{b, m}The values of registers 90 and 92 of (t) are first calculated. This includes the current , The next stage, and registers 90, 92 and Calculation of the intermediate variables and signal values used to determine the 96 values is required.   In block 180, Update mth order of regression filter (s) As shown, the value κ of the regression filter register 96 is then_{m, λa}(T) is calculated Is done. m is its predetermined maximum value (in this preferred embodiment, m = 6 or m = 10 ) Or indicated by the YES path from the "execute" decision of block 190. The two order update operations of blocks 170 and 180 until the values converge as The operation is executed m times in succession. In the computer subroutine, the weight of the forward prediction error Finding total F_m(T) and weight B of backward prediction error_m(T) is a small positive By checking if it is less than a number, convergence is determined. block The output is then calculated, as indicated by 200 "Calculate output". Figure 9 and the combined process estimating means 60 corresponding to the flowchart of FIG. This output is the primary or secondary signal, as determined by the subroutine Is a valid approximation to either Shown in block 210, Display This output is displayed (or summed by another subroutine) as Used in arithmetic).   Two measurement signals S_λa(T) and S_λbThe new sample set of (t) is shown in FIG. Of the processor and the joint process estimating means 60 corresponding to the flowchart of FIG. Is input to the is canceller subroutine, and processing is Will be restored. Note, however, that the initialization process is not performed again . Measurement signal sample S_λa(T) and S_λbThe new set of (t) is To the adaptive noise canceller subroutine. Is entered. The output forms a sample chain representing a continuous wave. This waveform Primary signal waveform s at wavelength λa_λa(T) or secondary signal waveform n_λa(T) A valid approximation to either. The waveform is also the primary signal wave at wavelength λb. Shape s_λb(T) and secondary signal waveform n_λbA valid approximation to any of (t) possible.   The flowchart corresponding to the QRD algorithm of FIG. This is shown in FIG.Calculation of saturation from output of correlation canceller   The physiologic monitor has a primary signal s "_λa(T) or s "_λb(T) or secondary credit No. n "_λa(T) or n "_λbUsing the approximation of (t), one or more components Calculate another quantity, such as the saturation of one component in a given volume containing the component. Generally, such calculations involve either primary or secondary signals at two wavelengths. We need information about the gap. For example, in the constant saturation method, the measurement signal S_{λ a} (T) and S_λbPrimary signal portion s of both signals of (t)_λa(T) and s_λb(T) A valid approximation is needed. The arterial saturation is an approximation for both signals, ie, s ″_λa( t) and s "_λb(T). The constant saturation method also applies to the secondary signal part n_λa(T) or n_λbRequires a valid approximation of (t). Vein satiety An estimate of the union is an approximation to these signals, ie, n ″_λa(T) and n "_λb(T ) Can be determined.   FIG. 10 shows a combined process estimator having two regression filters 80a and 80b. Step 60 is shown. The second regression filter 80b is used according to the constant saturation method. Measurement signal S_λb(T) to determine the reference signal n ′ (t) or s ′ (t) You. The first regression filter 80a and the second regression filter 80b are independent. Backward Prediction error b_m(T) is input to each of the regression filters 80a and 80b, An input to the filter 80b bypasses the first regression filter 80a.   The second regression filter 80b is arranged similarly to that of the first regression filter 80a. Register 98 and an adder element 108. The second regression filter 80b is Operates via additional intermediate variables with those defined by equations (54)-(64) . That is, ρ_{m, λb}(T) = λρ_{m, λb}(T-1) + ｛b_m(T) e^* _{m, λb}(T) / γ_m(T) ｝ (65) And ρ_{0, λb}(0) = 0 (66) It is. The second regression filter 80b calculates the error signal value e of the first regression filter._{m + 1, λa} It has an error signal value defined similarly to (t). That is, e_{m + 1, λb}(T) = e_{m, λb}(T) -κ^* _{m, λb}(T) b_m(T); (67) as well as, e_0,λ_b(T) = S_λb(T) When t ≧ 0 (68) It is. The second regression filter is defined similarly to the error signal value of the first regression filter. Regression coefficient κ_{m, λb}(T) The register has 98 values. That is, κ_{m, λb}(T) = ｛ρ_{m, λb}(T) / B_m(T)｝ (69) It is. These values are the intermediate variable values, signal values, and registers defined by equations (46) to (64). And register values. These signals are similar to the wavelength λa. Calculated at the order specified by placing an additional signal immediately next to the signal .   In the constant saturation method, S_λb(T) is input to the second regression filter 80b. This output is the primary signal s "_λb(T) or secondary signal s "_λbYes for (t) This is a valid approximation.   Even if the second regression filter 80b is added, it is shown by the flowchart of FIG. The computer program subroutine remains substantially unchanged. Only one regression Instead of updating the order of the mth stage of the filter, a regression filter 80a and The order of the m-th stage of both filters of 80b is updated. This is shown in FIG. Of the m-th stage of the regression filter (s) "Update degree" is characterized by multiple displays of filters. Regression filter Since 80a and 80b operate independently, the model Reference processor and combined process estimating means 60 Independent calculations can be performed in the sumbu routine.   The result of FIG. 10 using the QRD algorithm and using two regression filters Another diagram of the combined process estimator is shown in FIG. 10a. This type of join process The estimating means is a correlation cap using the QRD algorithm described in the Hikin reference. Used for cancel.Calculation of saturation   Valid approximation s "for the primary or secondary signal part_λa(T) and s "_λb(T) or n "_λa(T) and n "_λb(T) is a combined process estimating means Once 60 is determined, for example, A_FiveAnd A₆A in the volume containing_FiveSaturation of The degree can be calculated according to various known methods. Arithmetic, approximation to primary signal Is described using λa and λb as follows: s "_λa(T) ≒ ε_{Five, λa}c_Fivex_5,6(T) + ε_{6, λa}c₆x_5,6(T) (70) as well as, s "_λb(T) ≒ ε_{Five, λb}c_Fivex_5,6(T) + ε_{6, λb}c₆x_5,6(T) (71) Equations (70) and (71) provide three unknown values, c_Five(T), c₆(T) and x_5,6(T )). A_FiveAnd A₆A in the volume containing_FiveDegree of saturation And A_ThreeAnd A_FourA in the volume containing_ThreeThe saturation of which does not substantially change Different but close time t₁And t_Two, The primary signal part or the secondary signal The saturation can be determined by obtaining an approximation to the part. For example, for primary signals Then time t₁And t_TwoIs estimated as follows: Then, different signals related to the signals of equations (72)-(75) may be determined. That is: Here, Δx = x_5,6(T₁) -X_5,6(T_Two). Time t = (t₁+ T_Two) / 2 The average saturation at is: It is. It is understood that the Δx term has been omitted from the saturation calculation due to the division. Therefore To calculate the saturation, it is not necessary to know the thickness related to the primary component.Pulse oximetry   Using the processor of the present invention, the secondary signal portions that have been triggered Physiological monitor that determines the secondary criterion n '(t) that is input to the correlator canceller A specific example of a parameter is pulse oximetry. Using the processor of the present invention, Primary signal that can be used to play or input to the correlation canceller Signal criterion s' (t) to provide information on patient activity and venous blood oxygen saturation. Outgoing pulse oximetry may also be performed.   Typically, pulse oximetry is performed, for example, on the earlobe, fingers, fins, or fetal scalp. It propagates energy through the medium through which blood flows near such surfaces. Enel After the energy propagates through or is reflected from the medium, the attenuation signal is measured. pulse Oxygen measurement estimates the saturation of oxygen-containing blood.   Fresh oxygenated blood flows high from the heart to the arteries for use by the body. Sent out by pressure. Blood volume in the arteries changes with heart rate, and this change is Or a change in energy absorption in the pulse.   Depleted or deoxygenated blood can be used together with unused oxygenated blood. Returned to the heart by veins. Intravenous blood volume changes with respiratory rate, and respiratory rate It is typically much slower than your heart rate. Therefore, the vein thickness changes If not, venous blood causes low frequency changes in energy absorption. vein Low frequency changes in energy absorption when the thickness of Are coupled with the concomitant change in energy absorption due to motion artifacts.   In absorption measurement using energy propagation in a medium, blood flows near a surface such as a finger. Two light emitting diodes (LEDs) on one side of the body part Is provided with a photodetector. Typically, in pulse oximetry, one LE D emits a visible wavelength, preferably the red light wavelength, and the other LED emits an infrared wavelength. I do. However, those skilled in the art will recognize that other wavelength combinations may be used. Will understand. Fingers include skin, tissue, muscle, arterial and arterial blood, fat, etc. Each of which has different absorption coefficients, different concentrations, different thicknesses, and changes in optical path length. Absorbs light energy differently depending on Blood if the patient is not moving Except for the flow, the absorption is substantially constant. Constant through traditional filtering technology The attenuation is determined and subtracted from the signal. When the patient works, the background flow Optical path length due to body movement (eg, venous blood with a different saturation than arterial blood) Change occurs. Therefore, the measurement signal is adjoint. Incidental The noise generated by the operation is typically pre-determined via conventional filtering techniques. Cannot be determined and / or cannot be subtracted from the measurement signal No. Therefore, it becomes more difficult to determine the oxygen saturation of arterial and venous blood .   Schematic diagrams of a physiological monitor for pulse oximetry are shown in FIGS. FIG. FIG. 2 shows a general hardware block diagram of a pulse oximeter 299. Sensor 300 is L It has two light emitters 301 and 302 such as ED. LE that emits red wavelength light D301 and another LED 302 emitting infrared wavelength light are located adjacent to the finger 310 You. Generates electrical signals corresponding to attenuated visible and infrared light energy The resulting photodetector 320 is located on the opposite side of the LEDs 301 and 302. Light detection The device 320 is connected to the pre-analog signal conditioning circuit 330.   The pre-analog signal adjustment circuit 330 is connected to the analog-digital conversion circuit 332. Connect the output. The analog-digital conversion circuit 332 performs digital signal processing. The output is connected to the management system 334. Digital signal processing system 334 is It provides the desired parameters as output to display 336. For example, day Outputs for spraying include blood oxygen saturation, heart rate, and clear plethysmography. Waveform.   The signal processing system also supplies the digital-to-analog conversion circuit 338 with an emitter voltage. A current control output 337 is provided, and the digital-to-analog The control information for the driving circuit 340 is provided. The light emitter driving circuit 340 controls the light Connected to emitters 301 and 302. The digital signal processing system 334 is Also, a gain control output 342 for the pre-analog signal conditioning circuit 330 is provided.   FIG. 11a shows an emitter driving circuit 340 and a digital-analog conversion circuit 338. 3 shows a preferred embodiment of the combination of. As shown in FIG. 11a, the driving circuit First input latch 321, second input latch 322, synchronization latch 323, voltage reference 324, digital-analog conversion circuit 325, first switch bank 326, Two switch bank 327, first voltage-current converter 328, second voltage-current converter 329 and the LED emitters 3 corresponding to the LED emitters 301 and 302 in FIG. 01 and 302.   The preferred drive circuit shown in FIG. The present invention has found that most of the It is useful in that. Therefore, the emitter driving circuit of FIG. It is designed to minimize the noise from 302. A first input latch 321 and The second input latch 324 is connected directly to the DSP bus. So these latches Is the bandwidth present on the DSP bus leading to the drive circuit of FIG. ) Is considerably minimized. The outputs of the first input latch and the second input latch are It only changes when the latches detect their address on the DSP bus. The first input latch sets the digital-to-analog conversion circuit 325. receive. The second input latch is a switch for switch banks 326, 327. Receiving control data. The synchronization latch operates and activates the emitters 301 and 302. Receiving a synchronization pulse to maintain synchronization of the operation of the analog-to-digital conversion circuit 332 Take away.   The voltage reference also provides a low noise DC to digital-to-analog conversion circuit 325. Selected as voltage reference. Furthermore, in this embodiment, the voltage reference is very low Low-pass output filter with corner frequency (1 Hz in this embodiment, for example) Data. The digital-to-analog converter 325 also has a very low corner circumference. It has a low pass filter with a wave number (eg 1 Hz) at its output. Digital The analog-to-analog conversion circuit provides a signal to each of the emitters 301, 302. You.   In the present embodiment, the outputs of the voltage-current conversion circuits 328 and 329 are in a folded configuration. Only one of the emitters 301, 302 to be connected may be Switched to work on time. In addition, the voltage-current conversion circuit Switch off its input to the mitter so that it does not operate completely. This reduces noise from the switching circuit and the voltage-current conversion circuit. . In the present embodiment, a low noise voltage-current conversion circuit is selected (for example, Op 27 Op A mps), feedback to have a low pass filter to reduce noise A group is formed. In the present embodiment, as described later, the voltage-current conversion circuit The low pass filtering function of the paths 328 and 329 It has a corner frequency slightly higher than 625 Hz which is the switching speed. Therefore, FIG. The preferred drive circuit of a minimizes the noise of the emitters 301, 302.   Generally, each of the red light emitter 301 and the infrared light emitter 302 has energy And this energy is absorbed by the finger 310 and received by the photodetector 320. Taken. The photodetector 320 emits electrons corresponding to the intensity of light energy impinging thereon. Provide a signal. The pre-analog signal conditioning circuit 330 receives the intensity signal and These signals are adjusted for filtering as described. The signal obtained is The digital signal processing system 3 is provided to the analog / digital conversion circuit 332. 34 from an analog signal to a digital signal for further processing. You. Digital signal processing system 334 is referred to herein as "saturation conversion." Two signals are used to provide the conversion. Other than blood saturation monitoring For parameter monitoring, the saturation conversion depends on the desired parameters. It should be understood that concentration conversion, in vivo conversion, and the like are better. From the following description As will become apparent, the term saturation transformation is used to change the saturation The operation of converting the sample data into a threshold will be described. This embodiment , The output of the digital signal processing system 334 is Provide morphographic waveforms and provide oxygen saturation and pulse rate values to display 336 You.   It has been understood that different embodiments of the invention can provide one or more outputs. No. Digital signal processing system 334 also provides an emitter current control output 337. For driving the light emitters 301 and 302 together with the emitter current control signal at Provides control. This value is a digital value. The conversion circuit 338 converts the control signal to the emitter current drive circuit 340. Provide an issue. The emitter current drive circuit 340 includes a red light emitter 301 and an infrared light Providing a suitable current drive for the emitter 302. Physiological monitor operator Further details regarding the options are provided below. In this embodiment, the optical The driver is driven via the emitter current drive circuit 340 and is driven at 625 Hz Providing modulated optical transmission. In the present embodiment, the light emitters 301 and 302 It is detected by the detector and adjusted by the pre-analog signal adjustment circuit 330. Driven at power levels that provide acceptable strength. For a given patient, Once this energy level is determined by the digital signal processing system 334, The current levels of the red and infrared light emitters are kept constant. But However, in the room that affects the voltage input to the analog signal conditioning circuit 330, It should be understood that the current cannot be adjusted for changes in ambient light or other changes. Book In the invention, the red and infrared light emitters are modulated as follows: For one complete 5 Hz red light cycle, the red light emitter 301 の 間 cycle activated and the remaining ／ cycle turned off; 625 Hz For one complete infrared light cycle, the infrared light emitter 302 has a 1/4 cycle. And the remaining 3/4 cycle is turned off. Receive only one signal at a time For this reason, each of these emitters is 1/4 of a 625 Hz per cycle. It is operated only during the cycle, and is turned on and off alternately at intervals of 1/4 cycle.   Blood (or other sample medium) is pumped through finger 310 As a result, the optical signal is attenuated (amplitude modulated). Attenuated (amplitude modulated ) The signal is detected by the photodetector 3 at a carrier frequency of 625 Hz for red and infrared light. 20. Since only one photodetector is used, the photodetector 320 Receiving both the red light signal and the infrared light signal to form a composite time division signal You.   The composite time division signal is provided to a pre-analog signal conditioning circuit 330. Prefix The analog signal adjustment circuit 330 and the analog-to-digital conversion circuit 332 Further details are shown in FIG. As shown in FIG. 12, the pre-circuit 302 includes a preamplifier. 342, high pass filter 344, amplifier 346, programmable gain amplifier 348, and a low-pass filter 350. The preamplifier 342 is a photodetector 320 to convert the composite current signal into a corresponding voltage signal and amplify the signal. It is a simpedance amplifier. In the present embodiment, the amplifier simplifies the processing. To amplify the signal amplitude for the purpose. In the present embodiment, the auxiliary amplifier 34 The power supply voltages of No. 2 are -15 VDC and +15 VDC. As can be seen, the decay The signal shows ambient light as well as components that show infrared or red light depending on the time Including components. When light other than red light and infrared light exists near the sensor 300, Ambient light is detected by the light detector 320. Therefore, the gain of the preamplifier is Ambient light in the signal does not saturate the preamplifier under accurate and reasonable operating conditions To be selected.   In this embodiment, the preamplifier 342 is an AD74 manufactured by Analog Devices. Includes 3JR OpAmp. This transimpedance amplifier is suitable for the desired system. And some desirable features, such as; low equivalent input voltage noise, low equivalent input current noise. Noise, low input bias current, high gain bandwidth product, low total high frequency distortion, high common mode Exhibits rejection, high open loop gain, and high power rejection ratio It is particularly useful in that   The output of the preamplifier 342 is connected as an input to a high-pass filter 344. You. The output of the preamplifier is also supplied to an analog-to-digital conversion circuit 332 at a first input. 346 are provided. In the present embodiment, the high-pass filter has a frequency of about 1 / 2-1 Hz. It is a single pole filter with a corner frequency. However, in some embodiments, The corner frequency is easily increased to about 90Hz. As you can see, the red light signal 625Hz carrier frequency of signal and infrared light signal is more than 90Hz corner frequency Is also much higher. High pass filter 344 is coupled to amplifier 346 as input. Connect the output. In this embodiment, amplifier 346 includes a unit gain amplifier . However, the gain of the amplifier 346 depends on the variation of the single resistor. Adjustable. The gain of the amplifier 346 depends on the gain of the It is increased when it is reduced to compensate for the effect.   The output of amplifier 346 provides an input to programmable gain amplifier 348 . Programmable gain amplifier 348 also controls the gain on gain control signal line 343. A programming input from a digital signal processing system (DSP) 334 is received. The gain of the programmable gain amplifier 348 is digitally programmable. is there. The gain is initialized or the sensor configuration is changed because the test medium changes for each patient. Is dynamically adjusted. For example, the signals from different fingers are somewhat different. Therefore, Dynamically adjusted by programmable gain amplifier 348 to obtain a logically appropriate signal. A tunable amplifier is provided.   The programmable gain amplifier also has a separate emitter drive current that is held constant. It is also useful in the embodiment. In this embodiment, the analog-to-digital conversion To obtain the proper dynamic range at the input of circuit 332, the emitter drive The kinetic current is adjusted for each patient. However, changing the emitter drive current Changes the wavelength of the emitter, which in turn negatively affects the results of the oxygen measurement calculations. May be Therefore, the emitter drive current should be fixed for all patients. Will be useful. In another embodiment of the present invention, an analog-to-digital conversion circuit is provided. To obtain a signal that is properly within the dynamic range at the input, The ramable gain amplifier can be adjusted by the DSP. Thus the emitter The drive current is fixed for all patients and the wavelength shift due to changes in the emitter drive current. Can be reduced.   The output of programmable gain amplifier 348 is the input to low pass filter 350 Connected as Advantageously, low-pass filter 350 is 1 in this embodiment. It is a single-pole filter having a corner frequency of 0 kHz. This low pass filter Provides anti-aliasing in this embodiment.   The output of the low-pass filter 350 is sent to an analog-to-digital conversion circuit 332. A second input 352 is provided. FIG. 12 also shows the addition of an analog-digital conversion circuit. Shows additional disadvantages. In the present embodiment, the analog-digital conversion circuit 332 includes: First analog-digital conversion circuit 354 and second analog-digital conversion circuit 356. Advantageously, the first analog-to-digital conversion circuit 354 includes a first An input from the input 346 to the analog-to-digital conversion circuit 332 is received, and a second The analog-digital conversion circuit 356 is connected to the analog-digital conversion circuit 332. An input is received at a second input 352.   In one advantageous embodiment, first analog to digital conversion circuit 354 is diagnostic. Analog-digital conversion circuit. (By digital signal processing system The diagnostic work performed is performed when the input to the high-pass filter 344 is signal saturated. The output of the detector amplified by the preamplifier 342 to determine whether Is to read. In the present embodiment, the input to the high-pass filter 344 is saturated. When the condition occurs, the pre-analog signal conditioning circuit 330 provides a “0 (zero)” output. You. Alternatively, the first analog-to-digital conversion circuit 354 remains unused .   The second analog-digital conversion circuit 352 adjusts the signal from the pre-signal adjustment circuit 330. Receiving the trimmed composite analog signal and converting the signal to digital form. Book In an embodiment, the second analog-to-digital conversion circuit 356 includes a single channel Includes a Le-Sigma conversion circuit. In the present embodiment, CrystalSemicond uctor CS5317-KS delta-sigma analog-to-digital conversion Use a circuit. Such a conversion circuit is low cost and exhibits low noise characteristics Is useful in that More specifically, the delta-sigma conversion circuit comprises a noise modulator and And the decimation filter. The selected conversion circuit is , A second order analog delta-sigma modulator to provide noise shaping use. Noise shaping means that from flat response to higher frequency The noise at the lower frequencies by increasing the It refers to changing the noise spectrum. Then decimation (decimati on) The filter is reshaped to provide 16-bit performance at lower frequencies. High-frequency noise that has been cut. This conversion circuit generates 16 bits Data is sampled 128 times for each data word. Thus, the conversion circuit Provides excellent noise rejection, dynamic range and low frequency distortion, This is useful for critical measurement situations such as low perfusion and electrocautery.   Further, by using a single channel conversion circuit, two or more channels can be exchanged. There is no need to harmonize. The delta-sigma converter also enhances noise control Therefore, it is advantageous in exhibiting noise shaping. An example of analog-digital Tal conversion circuit is CS531 manufactured by Crystal Semiconductor 7 In the present embodiment, the second analog-digital conversion circuit 356 includes 20 The signal is sampled at a sample rate of kHz. 2nd analog-digital The output of the conversion circuit 356 converts the data samples at 20 kHz to a digital signal processing system. Provided on stem 334 (FIG. 11).   FIG. 13 shows the digital signal processing system 334 in detail. In this embodiment, , Digital signal processing system is microcontroller 360, digital signal Processor 362, program memory 364, sample buffer 366, data It includes a memory 368, a read-only memory 370, and a communication register 372. This implementation In an embodiment, the digital signal processor 362 is an analog devices company. AD21020. In the present embodiment, the microcontroller 360 Includes Motorola 68HCO5 built in program memory . In the present embodiment, the sample buffer 366 is 20 kHz sampled data from the data memory 36 8 is stored. In the present embodiment, the data memory 368 is a static random 32,000 words of the access memory (the word is 40 bits in this embodiment) Including).   The microcontroller 360 has a DSP via a conventional JTAG tap line. 362. The microcontroller 360 controls the program through the tap line. The boot loader for the DSP 362 is transferred to the program memory 364 and the DSP 3 62 is booted from the program memory 364. In the program memory 364 The boot loader in the read only memory 370 reads the operation instruction to the DSP 362. To the program memory 364. Advantageously, the program memory 3 64 is a very high speed memory for the DSP 362.   The microcontroller 360 controls the emitter current control via the communication register 372. Control and gain control signals.   14 to 20 illustrate the pulse rate performed by the digital signal processing system 334. It is a block diagram which shows the function of the operation of the oximeter 299. Signals to be described later The processing function is executed by the DSP 362 in the present embodiment, LA 360 provides system management. In this embodiment, the operation is software Hardware / firmware controlled. FIG. 14 shows the digital signal processing system 3 General operation performed on 20 kHz sample data input to 34 FIG. 4 is a block diagram showing typical functions. As shown in FIG. 14, the demodulation module 4 Demodulation indicated by 00 is first performed. Next, the decimation module 402 Is performed on the obtained data. Decimation In the data obtained from the operation, A predetermined statistic is performed as shown in FIG. The conversion is performed. Statistical operation performed data and saturation conversion operation The data that has been subjected to the saturation is represented by the saturation calculation module 408. Operations and pulse rate operations as indicated by the pulse rate calculation module 410 Sent to   In general, the demodulation operation separates the red and infrared signals from the composite signal. , 625 Hz carrier frequency, leaving raw data points. Decimation Raw data points are provided at 625 Hz intervals And the sample is 62. Reduced by a factor of 10 to 5 Hz samples. Decimeter The operation also provides some filtering on the sample. Statistical operations and saturation conversion operations are performed on the obtained data. Saturation, which is very resistant to motion artifacts and other noise in the signal The value is calculated. The saturation value is checked in the saturation calculation module 408 and the pulse Pulse rate and clear plethysmographic waveforms are obtained via the pulse rate module 410 You. For further details regarding the various operations, see FIGS. It will be described using FIG.   FIG. 15 shows the operation of the demodulation module 400. Demodulated signal format The result is shown in FIG. FIG. 15 shows one complete 625 Hz cycle of the composite signal And the first quarter cycle is the active red light and ambient light signal, 2 The first quarter cycle is the ambient light signal, and the third quarter cycle is active. The infrared signal and the ambient light signal, and the fourth quarter cycle is the ambient light signal. You. As shown in FIG. 15, when the sampling frequency is 20 kHz, A single full cycle at 625 Hz would yield 32 samples of 20 kHz data, ie, Eight samples related to red light and ambient light and eight samples related to ambient light 8 samples related to infrared and ambient light, and the final sample related to ambient light. 8 samples.   Since the signal processing system 334 controls the operation of the light emitters 300, 302, The whole system is synchronized. Data is demultiplexed (dMUX) module Using the time division demultiplex operation shown at 421, four 8 Simultaneously split (and thereby demodulated) into sample packets. One Eight sample packet 422 indicates the red light signal and the ambient light signal; the second eight samples Packet 424 shows the ambient light signal; third 8 sample packet 426 is attenuated Shows the infrared signal and the ambient light signal obtained; 3 shows an ambient light signal. The selection signal controls the demultiplex operation synchronously. At the input of the demultiplexing circuit 421, Divide into four subparts.   As shown in the add operation 430, 432, 434, 436 of FIG. As such, the sum of the last four samples from each packet is then calculated. This implementation In the embodiment, the last four samples are used in the analog-to-digital converter of the present embodiment. This is because the low-pass filter in the conversion circuit 356 has a set time. late Then, by collecting the last four samples from each eight-sample packet, The previous signal is clean. This addition operation is a product that enhances noise rejection. Provides arithmetic operations. As shown by the subtractor modules 438, 440 Then, if the sum of each ambient light sample is the sum of the red and infrared samples Is subtracted from the Subtraction operations reduce the amount of ambient light signal present in the data. Or attenuate. In the present embodiment, the operation of the subtraction modules 438 and 440 is performed. Provides about 20 dB of ambient light attenuation. Of the obtained red light and infrared light The sum is divided by four, as shown in the divide-by-4 modules 442,444. . Each value obtained is one sample of each of the 625 Hz red light signal and infrared light signal. I will provide a.   625 Hz carrier frequency removed by demodulation operation 400 Please understand that. 625 Hz support at the output of demodulation operation 400 The sample data is sample data without a carrier frequency. Nyquist sun To meet the pulling requirements, the human pulse is about 25-250 beats per minute, Or about 0. Less than 20Hz is necessary (understanding that it is 4Hz-4Hz) is there. Therefore, the resolution of 625 Hz in the decimation operation is 6 2. Reduced to 5 Hz.   FIG. 16 shows the operation of the decimation module 402. Red light The sample data of the red light signal and the infrared light signal Provided at 625 Hz to the buffer / filter 450, 452. In this embodiment, The red light signal and infrared light signal buffers / filters are 519 sample deep . Advantageously, buffer filters 450, 452 are continuous first in first out buffers. Function as a webcam. 519 samples are low-pass filtered It is. Preferably, low pass filtering is about 7 dB with about -110 dB attenuation. . It has a cut-off frequency of 5 Hz. Buffer / filter 450, 452 is 5 Form a finite impulse response (FIR) with coefficients for 19 taps You. To reduce the sampling frequency to 1/10, use the red light signal decimation module. 454 and infrared light decimation module 456, as shown in FIG. The calculation of the bandpass filter is performed every ten samples. In other words, each 10 When new samples are transferred to buffers / filters 450, 452, By multiplying the pulse response (coefficient) by 519 filter taps, A pass filter calculation is performed. By calculating each filter, the red light output buffer 4 58 and an output sample of the infrared light output buffer 460 are provided. In this embodiment The red light output buffer 458 and the infrared light output buffer 460 also It is also a continuous FIFO buffer that holds data samples. 570 Sump The infrared light sample and the red light sample or sample packet (in the text) Provides a “snapshot”). As shown in FIG. The buffer includes the statistical operation module 404 and the saturation conversion module 40 6, and sample data for the pulse rate module 410.   FIG. 17 shows the functional operation of the statistics module 404 in more detail. I have. In summary, the statistics module 404 includes a red light channel and an infrared light channel. Provide first order oximetry calculations and RMS signal values for the Statistics module It also provides a cross-correlation output that indicates the cross-correlation between the red light signal and the infrared light signal.   As shown in FIG. 17, the statistical operation removes the carrier frequency. Sample packets showing the attenuated infrared light signal and the red light signal (in this embodiment, Is 62. 570 samples at 5 Hz). Infrared light signal and red light Each packet of the signal is logged, as indicated by log modules 480,482. Is normalized using the function. After normalization, the DC removal modules 484, 486 The DC portion of the signal is removed, as shown by. In this embodiment, DC removal is Sample from each of the red and infrared light snapshots. Check the first one DC value and use this from all samples in each packet. Removing the DC value of.   When the DC signal is removed, the red light bandpass filter module 488 and the infrared light As shown by the optical bandpass filter module 490, the signal is bandpass filtered. Taling is performed. In this embodiment, each packet has 570 samples In some cases, the bandpass filter has a linear phase response and almost no distortion With 301 taps to provide a FIR filter with no or no It is composed of In this embodiment, the band-pass filter has a frequency of 34 beats / minute to 250 beats. It has a passband per minute per minute. 270 filters showing the filtered red light signal 270 filter processing showing the filtered sample and the filtered infrared light signal Tap on 301 slides over 570 samples to get a working sample . In the ideal case, bandpass filters 488 and 490 remove DC in the signal I do. However, the DC removal operations 484, 486 are not DC removal.   Module 492,4 selecting the last 120 samples after filtering As shown at 94, from each packet (of 270 samples in this embodiment) The last 120 samples are selected for further processing. Described below As described above, in the present embodiment, the saturation transfer module for processing the same data packet is used. Since the first 150 samples fall within the set time for the rule 406, the last 1 Twenty samples will be selected.   Conventional saturation equation calculation is applied to 120 sample packets of red light signal and infrared light signal. Be executed. In this embodiment, the conventional saturation calculation is performed in two different ways. You. One calculation is the first red light signal RMS module 496 and the infrared light signal RM As shown by the S module 498, the 120 sample packet is the entire RMS value Is processed to obtain The obtained RMS values of the red light signal and the infrared light signal are RED of 1 RMS / IR Provide input values for RMS ratio operation 500 Operation 500 includes an RMS as input to saturation equation module 502. It provides the ratio of the red light signal value to the RMS infrared light signal value. As will be appreciated by those skilled in the art. In addition, known red and infrared light wavelengths (typically λ_red= 650 nm and λ_IR= 910 nm) is the ratio of the intensity of the red light to the infrared light detected is the oxygen saturation of the patient. is connected with. Therefore, the saturation equation module 502 has a location at its output 504. 2 shows a conventional look-up table or the like that provides a known saturation value for a constant ratio. The red light RMS value and the infrared light RMS value are also determined by the saturation operation module 4. 04 is also provided as output.   In addition to the conventional saturation operation 502, the first cross-correlation module 506 A cross-correlation operation is performed on the 120 sample packets as shown. You. The first cross-correlation module 506 provides a good signal between the infrared light signal and the red light signal. Determine if a correlation exists. This cross-correlation can be defective or dysfunctional. It is useful to detect the detector of The cross-correlation is also the signal model (ie, equation (1) ) To (3) are also useful for detecting when the conditions are satisfied. Between two channels If the correlation between them is too low, the signal model will not be satisfied. To determine this, Normalized cross-correlation is performed on the cross-correlation module for each data snapshot. 506. One such correlation function is: Is shown.   If the cross-correlation is too low, the oximeter 299 alerts the operator (eg, To be visible, visible, etc.). In the present embodiment, the selected snack is If the snapshot gets a normalized correlation of less than 0.75, Do not qualify. Signals that meet the signal model have a correlation greater than the threshold .   120 sample packets of red and infrared light also have 5 120 samples In the same manner as described above, except that it is divided into sample bins of equal An operation and a cross-correlation operation are performed. RMS, ratio, saturation , And cross-correlation operations are performed on a bin-by-bin basis. These operations Are the five-segment equivalent bin modules 510 and 512 in FIG. Infrared light RMS modules 514, 516, second RED-RMS / IR-RMS ratio A module 518, a second saturation equation module 520, and a second cross-correlation module 522.   FIG. 18 shows further details regarding the saturation conversion module 406 shown in FIG. Show details. As shown in FIG. 18, the saturation conversion module 406 Sessa 530, correlation canceller 531, master power curve module 554, and And a bin power curve module 533. The saturation conversion module 406 is based on FIG. 7a having a quasi-processor 26, a correlation canceller 27, and an integrating means 29. In relation to this, one power curve is provided for the separate signal coefficients shown in FIG. 7c. be able to. The saturation conversion module 406 converts the data snapshot Obtain a saturation spectrum. In other words, the saturation conversion 406 is a snapshot Provides information on the saturation values present in the.   As shown in FIG. 18, the reference processor 53 of the saturation conversion module 406 is used. 0 is the reference generation module 534, the DC removal module 536, and the bandpass filter. It has a filter module 538. Red light from decimation operation And 570 sample packets of infrared light are provided to reference processor 530. In addition, a plurality of possible saturation values (“saturation axis scan”) are stored in the saturation reference processor 5. Provided as input to 30. In this embodiment, the saturation value of 117 is the saturation axis. Provided as a scan. In a preferred embodiment, the range of saturation values for 117 is 34. It evenly ranges from 8 to 105.0 blood oxygen saturation. Therefore, in this embodiment, 117 Is used as a reference signal for generating a reference signal used by the correlation canceller 531. An axis scan for the processor 530 is provided. In other words, the reference processor Provided a saturation value, so that a reference signal is generated corresponding to the saturation value . In the present embodiment, the correlation canceller includes the joint process estimating unit 550 and the low-pass The filter 552 is formed.   Scan values are chosen to provide higher or lower resolution than 117 scan values. Can be selected. The scan values may be unevenly spaced.   As shown in FIG. 18, the saturation equation module 532 includes a saturation axis as an input. Receive the scan value and output the ratio r_nI will provide a. General description of FIGS. 7a to 7c , This ratio r_nCorresponds to the plurality of scan values outlined above. The saturation equation is Provide a known ratio r (red / infrared) corresponding to the saturation value received as input. Offer.   Ratio r_nIs a reference generation circuit 53 like a sample packet of red light and infrared light. 4 is provided as input to. The reference generation circuit 534 generates a red light or an infrared light. Ratio r to any of the samples_nAnd multiply the values by infrared light or red light, respectively. Subtract from samples. For example, in the present embodiment, the reference generation circuit Ratio to color light sample_nAnd subtract this value from the infrared sample. Get The value obtained is the output of the reference generation circuit 534. This operation is a saturation run Calculated for each of the thresholds (eg, 117 values are possible in this embodiment). . Thus, the data obtained is 117 reference signals for each of the 570 data points. Vector, which will be referred to hereinafter as the reference signal vector. You. This data can be stored in an array or the like.   In other words, the red light sample packet and the infrared light sample packet Red light measurement signal S having a portion s '(t) and a secondary signal portion n' (t)_red(T) and Infrared light measurement signal S_IR(T), the output of the reference generation circuit is A secondary reference signal n ′ (t) according to the signal model obtained, which is expressed as follows: n '(t) = S_ir(T) -r_nS_red(T)   In the present embodiment, the reference signal vector and the infrared light signal are transmitted to the reference processor 530. As an input to the DC removal module 536. Statistics module 404 DC removal modules 536, such as DC removal modules 484, 486 in Is the DC value of the first sample for each input (or the first (Average of some or all samples). The sample values obtained are , Bandwidth pass filter 538.   The bandwidth pass filter 538 of the reference processor 530 Performs the same type of filtering as the bandwidth pass filters 488, 490 of . Thus, the 570 samples that have undergone the bandwidth pass filtering result in: There are 270 samples remaining. The first output 542 of the bandwidth pass filter 538 Is a single vector consisting of 270 samples (this embodiment In the form, a filtered infrared light signal is represented). Therefore, the bandwidth pass filter 5 The data obtained at the second output 540 of 38 are 270 data points each. 117 reference signal vectors, which are the saturation reference processor 53 0 corresponds to each of the saturation axis scan values provided.   The red light sample and the infrared light sample packet are stored in the reference processor 530. Can be switched when used. In addition, DC removal module 536 and And the bandwidth pass filter module 538 inputs the data to the reference processor 530. Can be performed before being forced. Because it runs in the reference processor This is because the calculations performed are linear. This results in considerable economics of processing. I will.   The output of the reference processor 530 is a combination processor of the type described above with reference to FIG. It should be appreciated that the first and second inputs to the estimator 550 are provided. The first input to the combining process estimator 550 represents an infrared light signal in this embodiment. 270 sample packets. This signal consists of a primary signal part and a secondary signal part. Including. The second input to the combining process estimator is 11 of 270 samples each. 7 is a reference signal vector.   The combined process estimator also includes a lambda input 543, a minimum error input 544, and The cell configuration number input 545 is received. These parameters are well understood in the art Have been. Lambda parameters are often referred to as "forget" This is called the “reject parameter”. A lambda input 543 is provided to the combined process estimator. Control of the cancellation rate. In this embodiment, lambda is a low value such as 0.8. Is set to Signal statistics are not constant, so track with lower values Is improved. The minimum error input 544 is the first input to the joint process estimator 550. Provide an initialization parameter (conventionally known as an "initialization value"). Book In an embodiment, the minimum error value is 10^-6It is. With this initialization parameter, The process estimating means 500 is not divided by 0 (zero) at the time of initial calculation. . The cell number input 545 of the combining process estimating means 550 is Configure the number of cells. In the present embodiment, the saturation conversion operation 406 The number of cells is six. As understood in the art, for each sine wave, The process estimator requires two cells. 35 to 250 beats / minute If there are two sine waves, two heart rate sine waves and one noise sine wave In contrast, there are six cells.   Combining process estimating means 550 provides a second input 540 to correlation canceller 531 A plurality of provided reference signal vectors (in this embodiment, all 117 consecutive First input 542 to the correlation canceller based on each of the Receive the vector. Correlation cancellation causes each of the 117 reference vectors to Thus, one output vector is obtained. Each output vector is paired with the first input vector. Indicates information that the corresponding reference signal vectors do not have in common. The resulting output vector Is provided as an output to the joint process estimator and is a low-pass filter Module 552. In the present embodiment, the low-pass filter 552 Tap and 10 Hz corner circumference with 62.5 Hz sampling frequency And an FIR filter having a wave number.   The joint process estimating means 550 of the present embodiment sets 150 data points. Have a pause. Therefore, from each of the 270 point output vectors, the last 120 data The points are used for further processing. In the present embodiment, the output vector is And further divided into a plurality of bins of equal number of data points. FIG. As shown at 8, the output vectors are output from the master power curve modules 554 and 5. Provided to the split bin module 556. The five-segment bin module 556 Divide the force vector into five bins of equal number of data points. Each bin is then , Bin power curve module 558.   The master power curve module 554 performs the saturation conversion as follows: For each output vector, the sum of the squares of the data points is ascertained. this is, Corresponds to each output vector (each output vector corresponds to one of the saturation scan values) Provides the sum of the squared values. These values are used to determine the master power as shown in FIG. -Provides a base for curve 555. The horizontal axis of the power curve is the saturation axis scan value And the vertical axis indicates the sum of squared values (or output energy) of each output vector . In other words, as shown in FIG. 22, each sum of squares produces an output vector Vertical "energy" at a point on the horizontal axis of the corresponding saturation scan value Plotted using the magnitude of the sum of the squared values plotted on the "output" axis Can be. As a result, a master power curve 558 is obtained, an example of which is shown in FIG. Is shown in This considers all possible saturation values and assumes the assumed saturation value By examining the output value for A saturation conversion is provided. As will be understood, the correlation canceller 531 When the first input and the second input are almost correlated, the correlation canceller 5 The sum of the squares of the 31 corresponding output vectors is very low. Conversely, correlation canceller If the first and second inputs to 531 show little correlation, the output vector The sum of the squares is high. Therefore, the spectral content of the reference signal and the second One input is mostly physiological noise (venous blood movement due to breathing) and non-physiological Output noise (e.g., generated by operation) Ghee will be low. Reference signal spectral content and first input to correlation canceller If the forces are not correlated, the output energy will be higher.   The corresponding transform is the same except that a saturation transform power curve is generated for each bin. And is performed by the bin power curve module 558. The resulting power curve Is provided as an output of the saturation conversion module 406.   Generally, according to the signal model of the present invention, as shown in FIG. There are two peaks. One peak corresponds to arterial oxygen saturation of blood And another peak corresponds to the venous oxygen concentration of the blood. In the signal model of the present invention Relatedly, the peak corresponding to the highest saturation value (the peak with the largest magnitude) Need not be) is the proportionality factor r_aCorresponding to In other words, the proportional coefficient r_aIs dynamic It corresponds to the red / infrared light ratio measured for pulse saturation. Similarly, lowest saturation The peak corresponding to the degree value (not necessarily the peak with the lowest magnitude) is roughly Corresponds to the venous oxygen saturation, which is used in the signal model of the present invention. Proportional coefficient r_vCorresponding to Therefore, the proportional coefficient r_vCorresponds to venous oxygen saturation Red light / infrared light ratio.   To obtain arterial oxygen saturation, peak the power curve corresponding to the highest saturation value. You can choose. However, further processing is required to increase the reliability of the values. Is executed. FIG. 19 shows the saturation conversion module 406 and the statistics module 40. 4 illustrates the operation of the saturation calculation module 408 based on the output of FIG. FIG. As shown in FIG. 5, the bin power curve and the bin statistics are stored in the saturation calculation module 40. 8 is provided. In this embodiment, the master curve is provided to the saturation module 408. Rather, display for visual check in system operation Can be done. The bin statistics are the red light RMS value and infrared light RMS value, seed (Seed) the saturation values and the red and infrared light signals from the statistics module 404; And a value indicating a cross-correlation between them.   The saturation calculation module 408 is first displayed by the bin attribute calculation module 560. A plurality of bin attributes are determined as follows. The bin attribute calculation module 560 Data bins are collected from the information from the power curves and the information from the bin statistics. Book In an embodiment, this operation corresponds to the highest saturation value in the data bin. And arranging the saturation values of the peaks from each power curve. This embodiment State, by convolving the smoothed differential filter function and the power curve, By first calculating the first derivative of the power curve Be executed. In the present embodiment, a smoothing differential filter function (using an FIR filter Has the following coefficients: This filter performs differentiation and smoothing. Next, the target original Each point in the power curve is evaluated, and the following conditions (1) and (2) are satisfied. Is determined to be a possible peak if: (1) the point in the power curve At least 2% of the maximum; (2) the value of the first derivative is greater than 0 and less than or equal to 0 It changes to the number below. For each point found to be a possible peak, The contact points are examined and the largest of the three points is the true peak It is considered to be.   The peak width for these selected peaks is also calculated. Target power The peak width of the curve is calculated by summing all points on the power curve It is calculated by subtracting the product of the minimum value and the number of points. In this embodiment The peak width calculation is performed for each of the bin power curves. Maximum is peak Selected as width.   Further, the RMS value of the infrared light, the RMS value of the red light, The seed saturation value for the bin, and the red light signal from the statistics module 404 The cross-correlation with the infrared light signal is also located in the data bin. Then use attribute Data bins are acceptable as shown in bin qualification logic module 562 It is determined whether the data is composed of various data.   If the correlation between the red and infrared light signals is very low, the bin is discarded . Seed for bins with the same saturation value for the selected peak for a given bin If so, the peak is replaced with a seed saturation value. Red light Either the RMS value or the infrared RMS value is below a very small threshold If so, all bins are discarded and no saturation value is provided. Because the measurement signal is small This is because it is determined that it is not possible to obtain meaningful data. Bottle If the data does not contain allowable data, the exception processing module Is provided to the display 336.   If some bins qualify, they are qualified as having acceptable data Those bins that are not qualified are replaced with the average of the allowed bins. Each bin is time stamped to maintain a time sequence. baud Operation 565 examines each bin and selects the highest saturation value. This These values are sent to clip and smooth operation 566.   The clip and smooth operation 566 basically uses a low pass filter. And perform the averaging process. Low-pass filter, smoothing filter selection module Provide an adjustable smoothing process selected by 568. Smoothing filter selection Module 568 includes the reliability performed by high reliability test module 570. Perform the operation based on the decision. High reliability test is bin power It is an investigation of the peak width for the curve. The peak width is some kind of table that indicates patient activity. An indication is provided, indicating operation if the peak width is wide. Therefore, if the peak width is wide, The smoothing filter is slowed down. If the peak width is narrow, the speed of the smoothing filter The degree increases. Therefore, smoothing filter 566 adjusts based on the level of confidence. Is adjusted. The output of clip and smoothing module 566 is Provides the saturation value.   In this preferred embodiment, the clip and smoothing filter 566 uses a new saturation Takes a degree value and compares it to the current saturation value. The magnitude of the difference is 16 (% If it is less than (oxygen saturation), the value is acceptable. Instead, the new saturation value If it is less than the filtered saturation value, the new saturation value is filtered. It is changed to a value 16 lower than the controlled saturation value. New saturation value is filtered If it is greater than the filtered saturation value, the new saturation value will be The value is changed to a value larger by 16 than the degree value.   While high reliability is present (no operation), the smoothing filter is: Here is a single-pole or exponential smoothing filter computed as:     y (n) = 0.6^*x (n) +0.4^*y (n-1) Where x (n) is the new clipped saturation value and y (n) is the filter The processed saturation value.   During operation, a 3-pole IIR (finite impulse response) filter is used. It is. Its features have values of 0.985, 0.900, and 0.94 respectively. Three time constants t_a, T_b, And t_cIs controlled by 3 using the following relationship The coefficients for the form I, IIR filters are calculated directly from the two time constants: a₀= 0 a₁= T_b+ (T_c) (T_a+ T_b) a_Two= (-T_b) (T_c) (T_a+ T_b+ (T_c) (T_a)) a_Three= (T_b)^Two(T_c)^Two(T_a) b₀= 1-t_b− (T_c) (T_a+ (T_c) (T_b)) b₁= 2 (t_b) (T_c) (T_a-1) b_Two= (T_b) (T_c) (T_b+ (T_c) (T_a)-(T_b) (T_c) (T_a) -T_a)   20 and 21 show the pulse rate module 410 (FIG. 14) in more detail. As shown in FIG. 20, the heart rate module provides transient rejection and bandwidth Filter module 578, motion artifact suppression module 580, saturation Equation module 582, operating state module 584, first and second spectrum estimation Modules 586, 588, spectrum analysis module 590, slew rate control Limit module 592, output filter 594, and output filter coefficient module 5 96.   As further shown in FIG. 20, the heart rate module 410 includes a decimation mode. 570 sample snapshots of infrared and red light from the output of Joule 402 Receive The heart rate module 410 further includes a saturation calculation module 408 Receives the saturation value output from. Further, the reliability test module 570 The maximum peak width (same as the peak width calculation above) calculated by Provided as input to the beat rate module 410. Infrared and red light sample The output of the bracket, saturation value, and operating status module 584 are the operating artifacts. To the control module 580.   The average peak width value provides an input to the operating status module 584. This embodiment In an embodiment, if the peak width is wide, this is taken as an indication of activity. motion If no is detected, the spectral estimate in the signal will be Executed directly without.   If motion is detected, the motion artifact is set to the motion artifact suppression mode. It is suppressed using Joule 580. Motion Artifact Suppression Module 580 Is substantially the same as the saturation conversion module 406. Motion artifact suppression module Rule 580 is connected as an input to the second spectrum estimation module 588 Provide output. First spectrum estimation module 586 and second spectrum estimation Module 588 provides an output that provides an input to spectrum analysis module 590. Have. The spectrum analysis module 590 also includes a Take output as input. The output of the spectrum analysis module 590 is the heart rate Initial heart rate determination of module 410, slew rate limiting module 592 Provided input to The slew rate limiting module 592 controls the output filter 5 Connected to 94. Output filter 594 also includes output filter coefficient module 5 Input is received from 96. The output filter 594 is connected to the display 336 (FIG. 11). ) To provide the filtered heart rate.   If there is no operation, indicated by the DC reject and bandwidth filter module 578 As such, one of the signals is DC rejected and bandwidth filtered. DC removal and The bandwidth filter module 578 includes the DC removal and bandwidth filter module 5. 36, 538 provide the same filtering. During periods of inactivity, the filter The processed infrared light signal is provided to a first spectral estimation module 586.   In this embodiment, the spectrum estimation provides a frequency spectrum of the heart rate information. Chirp Z transformation. The frequency range for the desired output is Chirp Z instead of the traditional Fourier Transform Use transformations. Therefore, in the present embodiment, a heart rate of 30 to 250 beats / min. A wave number spectrum is provided. In the present embodiment, the frequency spectrum is To the spectrum analysis module 590, which selects the first harmonic from the Provided. Typically, the first harmonic has the largest magnitude and pulse rate Is a peak in the frequency spectrum. However, in some situations, the second The third harmonic or the third harmonic can exhibit a larger magnitude. You. With this in mind, to select the first harmonic, the largest peak in the spectrum The first peak having an amplitude of at least 1/20 of is selected. by this, It is most possible to select the peak of the chirp Z-transform caused by noise as the heart rate Be reduced.   If an action occurs, the action artifact suppression module 580 is used to scan the action. Motion artifact suppression is performed in a nap shot. Motion artifact The event suppression module 580 is shown in more detail in FIG. As shown in FIG. The product artifact suppression module 580 includes a saturation conversion module 406 (FIG. 1). It is almost the same as 8). Therefore, the motion artifact suppression module Artifact reference processor 570 and motion artifact correlation canceller 571 Having.   The motion artifact reference processor 570 includes a The reference processor 530 is the same. However, the reference processor 570 Rather than using the saturation scan value of 117 to perform the overall saturation conversion, Use the saturation value from module 408. Therefore, the reference processor 570 A saturation degree module 581, a reference generation unit 582, a DC removal module 583, And a bandwidth pass filter module 585. These modules are This is the same as the corresponding module in the sum degree conversion reference processor 530. Real truth In the embodiment, the saturation expression module 581 is included in the saturation conversion module 406. Instead of performing the saturation axis transformation as performed by An arterial saturation value is received from 408. Because arterial saturation was selected, This is because there is no need to perform axis scanning. Accordingly, the saturation degree module 581 The output is the proportional constant r_a(Ie red light to infrared light ratio for arterial saturation value) You. Otherwise, the reference processor 570 uses the base of the saturation conversion module 406. Performs the same functions as sub-processor 530.   The motion artifact correlation canceller 571 also includes a saturation conversion correlation canceller. 531 (FIG. 18). However, the motion artifact suppression correlation key The canceller 571 uses a slightly different motion artifact combination process estimator 5. Use 72. Therefore, the motion artifact suppression correlation canceller 571 is A combined process estimating unit 572 and a low-pass filter 573 are provided. Motion arche The cell combination process estimating means 572 selects the cell selected by the cell number input 574. Are different (6 to 10 in the present embodiment), and the forgetting parameters are different ( In this embodiment, 0.98), and the time delay differs depending on the adaptation. This is different from the sum conversion transformation process estimation means 550. The low-pass filter 573 is This is the same as the low-pass filter 552 of the sum degree conversion correlation canceller 531.   Since only one saturation value is provided to the reference processor, each of the 570 samples For the input packet, output to the operation artifact suppression correlation canceller 571 In this case, only one output vector of 270 samples is obtained. In this embodiment, If an infrared light wavelength is provided as the first input to the correlation canceller, the correlation The output of the canceller 571 provides a clear infrared light waveform. As mentioned above, infrared light waves The long signal and the red light wavelength signal are output from the operation artifact suppression correlation canceller 571. The output can be switched to provide a clear red light waveform. Wear. Reference processor 570 is input to correlation canceller 571 as a reference signal. Knowing the actual saturation value of the patient that can generate a noise reference The output of the canceller 571 has a clear waveform. Motion artifact suppression module 5 The clear waveform at the output of 80 is a clear pre-view that can be sent to the display 336. It is a chismograph.   As mentioned above, an alternative joint process estimator is a QRD least squares grid type method ( FIG. 8A, FIG. 9A and FIG. Therefore, the combining process estimating means 57 3 (as well as the combined process estimator 550) is a QRD least squares lattice type operation. Alternatively, it can be replaced by a combined process estimating means for executing an action.   FIG. 21A shows the combined process estimating means 572a Fig. 4 shows another embodiment of a replaced motion artifact suppression module. Join The process estimating means 572a is a QRD least squares grid type system as shown in FIG. including. According to this embodiment, different QRD algorithms may be used as needed. Initialization parameters are used.   The initialization parameters are “cell number”, “lambda”, and “minimum error total” in FIG. , “Initial γ”, and “initial error total”. The number of cells and lambda are Corresponding to similar parameters in the process estimating means 572. "Initial γ" B corresponds to the gamma initialization variable for all stages except the As described above, it is initialized to 1 in the QRD expression. The initial error sum is Provides the δ initialization parameters described above. QRD to avoid overflow The larger of the denominator actually calculated for each division in the equation and the minimum error total is used. Used. In this embodiment, the preferred initialization parameters are as follows:     Number of cells = 6     Lambda = 0.8     Minimum error sum = 10^-20     Initial γ = 10^-2     Initial error total = 10^-6   The clear waveform output from the motion artifact suppression module 580 also Provide input to the spectral estimation module 588. Second spectrum estimation module Module 588 performs the same chirp Z-transform as the first spectrum estimation module 586. Execute. If there is no operation, the output from the first spectrum estimation module 586 The force is provided to the spectrum analysis module 586. If an action occurs, The output from the two-spectrum estimation module 588 is the spectrum analysis module 59 0 is provided. Spectral analysis module 590 provides an appropriate spectral estimation module. The frequency spectrum from the module is examined to determine the pulse rate. When the action occurs If so, the spectrum analysis module 590 has the highest amplitude in the spectrum. Select the peak to be used. Because the motion artifact suppression module 580 is completely This is because all other frequencies are attenuated to values below the actual heart rate peak. motion If not, the spectrum analysis module will specify the heart rate as described above. Select the first harmonic in the torque.   The output of the spectrum analysis module 590 is output to the slew rate limiting module 59. Provide the raw heart rate as input to 2. In the present embodiment, the slew rate limiting mode is used. Joule 592 prevents changes of more than 20 beats / minute per 2 second interval. Stop.   Output filter 594 is described above with respect to clip and smoothing filter 566. Includes an exponential smoothing filter similar to an exponential smoothing filter. Output file The filter is controlled via output filter coefficient module 596. Large movement If not, this filter slows down and has little or no activity In some cases, this filter samples faster and keeps distinct values be able to. The output from output filter 594 is the patient's pulse, which is Provided to spray 336 and useful.Alternative to Saturation Conversion Module-Bank of Filters   An alternative method of the saturation conversion of the saturation conversion module 406 is shown in FIG. This can be done using a bank of data. As shown in FIG. 23, the first filter bank Two filter banks are provided, 600 and a second filter bank 602. No. One filter bank 600 has a second filter bank at a corresponding first filter bank input 604. 1 measurement signal S_λb(T) (in this embodiment, an infrared light signal sample) is received and the second Filter bank 602 receives a second measurement at corresponding second filter bank input 606. Constant signal S_λa(T) (in this embodiment, a red light signal sample) is received. Suitable In an embodiment, the first and second filter banks have a fixed center frequency and a fixed corner frequency. -Use a static regression polyphase bandpass filter with frequency. Regression polyphase fill Ta is described in Harris et al. This document `` Digital Single P rocessing With Efficient Polyphase Recursive All-Pass Filters '' A is attached. However, adaptive execution is also possible. This embodiment State, the regression polyphase bandpass filter elements each have a specific center frequency and And bandwidth.   There are N filter elements in each filter bank. 1st filter bar Each of the filter elements in link 600 , A filter that matches (ie, has the same center frequency and bandwidth) Has elements. As shown in FIG. 23, the center circumference of N elements The wave numbers and corner frequencies are each N frequency ranges, from 0 to F₁, F₁-F_Two, F_Two-F_Three, F_Three-F_Four... F_N-1-F_NDesigned to occupy. (206) The number of filter elements can range from 1 to infinity. Only However, in the present embodiment, the center frequency is 25 beats / minute to 250 beats / minute. Approximately 120 distinct filter elements to span the frequency range of .   The output of the filter is the first measurement signal and the second measurement signal (this Examples include information about primary and secondary signals (red light and infrared light). Ma Filters (one in the first filter bank 600 and the second filter bank) The output of each pair (one at link 602) is provided to saturation determination module 610. Provided. FIG. 23 shows one saturation determination module 610 for ease of explanation. Only show. However, saturation for each matching pair of filter elements A degree determination module is provided and parallel processing can be performed. Each saturation determination The module has a ratio module 616 and a saturation type module 619.   The ratio module 616 forms a ratio of the second output to the first output. For example, in this example Is the ratio of each red light RMS value to each corresponding infrared light RMS value (red light / infrared light) It is determined in Joules 616. The output of the ratio module 616 is Provide an input to a saturation equation module 618 that references the corresponding saturation value for the You.   The output of the saturation equation module 618 is for each of the matching filter pairs (As shown in the histogram module 620). However However, the data collected is initially a function of frequency and saturation. As shown in FIG. To form a saturation conversion curve similar to the curve A gram or the like is generated. The horizontal axis indicates the saturation value, and the vertical axis indicates the saturation value at each saturation value. Shows the total number of points collected (output from saturation equation module 618) . In other words, the saturation equation module for ten different filter matching pairs If the output of the rule 618 shows a 98% saturation value, the histogram of FIG. The points reflected reflect a value of 10 at 98% saturation. Thereby, in FIG. A curve similar to the saturation conversion curve results. This operation uses the histogram Executed in module 620.   The histogram results provide a power curve similar to the power curve of FIG. . Therefore, the arterial saturation is the peak corresponding to the highest saturation value (in the region of interest). The maximum number of occurrences (eg, the peak in the drawing corresponding to the highest saturation value peak) It can be calculated from the histogram by selecting c). Similarly, The vein or background saturation is calculated by the saturation calculation module 408. In a similar manner to the processing, the peak is selected by selecting the peak corresponding to the lowest saturation value. Can be determined from the gram.   An alternative to the histogram is to use the output saturation (histogram) corresponding to the highest saturation value. Need not be the peak in the ram), and r represents the corresponding ratio_aArterial saturation with The degree can be selected. Similarly, the output saturation corresponding to the lowest saturation value is And r representing the corresponding ratio_vVein with background or background saturation You can choose. For example, in the present embodiment, the entry in the histogram of FIG. Select a as the arterial saturation and select the entry in the histogram with the lowest saturation value. Select bird b as vein or background saturation.   As described above, according to the present invention, the primary The signal part and the secondary signal part can be modeled as follows.     S_red= S₁+ N₁                  (Red light) (89)     S_IR= S_Two+ N_Two                   (Infrared light) (90)     s₁= R_as_TwoAnd n₁= R_vn_Two                                    (91) When equation (91) is inserted into equation (89), the following equation is obtained.     S_red= R_as_Two+ R_vn_Two            (Red light) (92)   In the models of equations (89) to (92), S_redAnd S_IRUse It is This is because the description specifically relates to blood oximetry. S_redAnd S_IRIs the previous sentence S inside₁And S_Two, And the following description is based on the arbitrary measurement signal S₁And S_TwoAgainst Done.   As explained above, r_aAnd r_v(These are arterial blood oxygen and And venous blood oxygen), the saturation conversion described above, scanning many possible coefficients Can be determined. R based on red and infrared data_aAnd r_v Another way to get_kIs at least some (preferably substantially) n_kCorrelates with No (where k = 1 or 2), s_kAnd n_kCorrelation between R to minimize_aAnd r_vIs to look for. These values are the following for k = 2: It can be found by minimizing the statistical calculation function. In the formula, i indicates time.   It should be understood that other correlation functions can be used, such as normalized correlation.   If the noise component does not correlate with the desired signal component, minimizing this amount will increase In many cases a single pair of r_aAnd r_vIs provided. The minimization of this quantity is given by the equation S from (90) and (92)_TwoAnd n_TwoAnd r_aAnd r_vFind the minimum value of the correlation between the possible values of Can be achieved. s_TwoAnd n_TwoBy seeking the following Is provided. The 2x2 matrix provides:   Therefore, Or: Preferably, the correlation in equation (93) is a window specified by the user as follows: Augmented using functions.   The Blackman Window is the preferred embodiment here. is there. That there are many additional functions that minimize the correlation between signal and noise I want to be understood. The above function is one of simple functions. Therefore,   Red light sample to minimize at multiple discrete data points Sum of squared points, infrared light sample Sum of squared points, red light sample The sum of the product of the point and the infrared sample point is first calculated (window Number w_iincluding).   These values are used in the correlation equation (93b). Therefore, the correlation equation is given by two variables r_aas well as r_vIs an expression using r_aAnd r_vTo obtain_aAnd r_vValid values of A thorough scan is performed on the various cross sections. Then the minimum of the correlation function A value is selected, yielding the minimum value_aAnd r_vIs r_aAnd r_vIs selected as   r_aAnd r_vIs obtained, the ratio r_aAnd r_vStatistics that provide oxygen saturation corresponding to In the saturation expression such as the saturation expression 502 of the module 404, r_aAnd r_vTo provide Thus, the arterial oxygen saturation and the venous oxygen saturation can be determined.   r_aAnd r_vThe same signal model set as above in further runs to obtain Use According to this execution r_aAnd r_vTo determine the signal s_TwoEnergy inside Lugie is s_TwoIs n_TwoMaximized under the constraint that it does not correlate with Again, this The implementation is based on minimizing the correlation between s and n in the signal model of the present invention. Good. Where signal s is related to arterial pulsation and signal n is noise (motion artifacts and And information about venous blood as well as other noise); r_aIs an artery The ratio associated with the sum (red light / infrared light), r_vIs the ratio related to venous saturation (red Light / infrared light). Thus, in this implementation of the invention, r_aAnd r_vIs the signal s_TwoAnd n_TwoAre not correlated, the signal s_TwoDetermined to maximize energy Is done. Signal s_TwoEnergy (ENERGY (s_Two)) Is defined by the following equation: Where R₁Is the energy of the red light signal, R_TwoIs the energy of the infrared light signal, and R_1,2 Is the correlation between the red light signal and the infrared light signal.   s_TwoAnd n_TwoCorrelation (s_Two, N_Two)) Is defined by:   As explained above, the constraint is s_kAnd n_k(K = 2 in this embodiment) That is not. This “correlation cancellation constraint” is expressed by the equation (97) Is set to zero. -R₁+ (R_a+ R_v) R₁₂-R_ar_vR_Two= 0 (98) In other words, the goal is to maximize equation (94) under the constraints of equation (98).   To achieve the objective, a cost function is defined as follows (for example, Example Lagrange optimization). Where μ is a Lagrange multiplier. Solve the cost function r_a, R_vAnd μ are "Luenberger, Linear & Nonlinear Programming" (Addison-Wesley (Addison-Wesley), 2nd edition, 1984) Can be found using   Along the same line, the red light signal S_redAnd infrared light signal S_IRIs not static Contains the function R defined above₁, R_Two, And R₁₂Is time dependent. Therefore, two Using the equation, the correlation cancellation system defined by equation (98) can be obtained at two different times. By showing the approximation, two unknowns can be obtained. The correlation cancellation constraint is Two different times t₁And t_TwoCan be expressed as follows. -R₁(T₁) + (R_a+ R_v) R₁₂(T₁) -R_ar_vR_Two(T₁) = 0 (100) -R₁(T_Two) + (R_a+ R_v) R₁₂(T_Two) -R_ar_vR_Two(T_Two) = 0 (101)   Equations (100) and (101) are_aAnd r_vIs nonlinear at These two equations can be solved using linear techniques. Therefore, x = r_a+ R_v Y = r_ar_vEquations (100) and (101) are obtained using R₁₂(T₁) XR_Two(T₁) Y = R₁(T₁) (102) R₁₂(T_Two) XR_Two(T_Two) Y = R₁(T_Two) (103)   These equations (102) and (103) can be solved for x and y. Variable expression changes From r_aAnd r_vSolving for gives the following:   Equation (104) is r_vConsequently. In this embodiment, x^Two-R_vy> 0 R_vThe value is selected. r_vAre both x^Two-R_vIf y> 0, t_TwoIn S_TwoEnergy (s_TwoR) to maximize))_vIs selected. r_vIs next Given by the above equation,_aIs obtained. Or r_vIn the same manner as the determination of_a Can also be found directly.Alternative Method of Saturation Conversion-Complex FFT   The patient's blood oxygen saturation, pulse rate and clear plethysmographic waveforms also It is also obtained using the signal model of the invention using a plex (complex) FFT, This will be further described with reference to FIGS. 25A to 25C. Generally, each is the first A portion (indicating a desired portion of the signal) and a second portion (indicating an unnecessary portion of the signal). (The two measurement signals have coefficients r_aAnd r_vCan be correlated by Can be calculated by using the signal models of equations (89) to (92) From the output of the binary operation 402 at discrete sample points. A fast saturation transformation of the source can be used.   FIG. 25A substantially corresponds to FIG. 14, but the above-described saturation conversion is a fast saturation conversion. I am. In other words, the operation of FIG. 25A corresponds to the operation of FIG. Can be replaced. As shown in FIG. 25A, the fast saturation conversion This is indicated by the diversion / pulse rate calculation module 630. As in FIG. 14, the output is arterial oxygen. Saturation, clear plethysmographic waveform, and pulse rate. FIG. 25B and FIG. C shows further details about the fast saturation conversion / pulse rate calculation module 630. doing. As shown in FIG. 25B, the fast saturation conversion module 630 17 as the infrared light log module 480 and the red light log module 482. Infrared light log module 640 and red light log module 64 for performing log normalization 2 Similarly, the infrared light DC removal module 644 and the red light DC module 646 exists. Further, the infrared light high-pass filter module 645 and the red Color light high-pass filter module 647, window function modules 648, 6 40, complex FFT module 652, 654, selection module 653 , 655, magnitude modules 656, 658, threshold module 66 0, 662, point-by-point ratio module 670, saturation degree module 672, and And a selective saturation module 680. Phase module 690, 692 , A phase identification module 694, and a phase threshold module 696 are also present. Exist. The output of the selection saturation module 680 is connected to the arterial saturation output line 682. Provide arterial saturation.   In this alternative embodiment, snapshots of the red and infrared light signals are 562 samples from the simulation module 402. Infrared light DC removal The module 644 and the red light DC removal module 646 correspond to the infrared light DC Slightly different from removal module 484 and red light DC removal module 486 I have. The infrared light DC removal module 644 and the red light DC removal module of FIG. 25B 646, the average of all 563 sample points for each channel is summed. Is calculated. This average was then used to remove baseline DC from each sample. Removed from each individual sample point in each snapshot Is done. Infrared light DC removal module 644 and red light DC removal module 646 Are the infrared light high-pass filter module 645 and the red light high-pass filter module. An input is provided to each of the modules 647.   High pass filter modules 645, 647 have 51 taps of coefficients Includes FIR filter. Preferably the high pass filter has a side lobe of 30 Has a level parameter and a corner frequency of 0.5 Hz (ie 30 beats / min) Includes a Chebychev filter. Change the performance of this filter Please understand that it is possible. 562 sample points are high pass fill And there are 51 coefficient taps, so the output of the high-pass filter module Are the infrared and red light snapshots of each of these 512 samples are provided. The output of the high-pass filter is Inputs are provided to the window function modules 648, 650 for the channel.   Window function modules 648, 650 perform conventional window functions. You. In the present embodiment, a Kaiser window function is used. FIG. 25B Throughout the function maintains a point-by-point analysis. In this embodiment, the Kaiser window The time bandwidth product of the function is 7. The output of the window function module is Fast Fourier Transform (FFT) module 652, 654 provide input.   Complex FFT modules 652 and 654 provide data snapshots Complex in each of infrared and red light channels at Perform FFT. The data from the complex FFT then goes through two paths One of the paths explores magnitude and the other is compressed Investigate the phase from the FFT data points. However Before further processing, the data is stored in the infrared light selection module 653 and the red light selection module. Joule 655. Because the output of FFT operation is 0-1 Because it provides repetition information from a sampling rate of 1/2 and a sampling rate of 1/2 It is. The selection module has a sampling rate of 0 to （(in the present embodiment, for example, 0-31.25 Hz), select only the sample, the heart rate frequency range and heart rate. A selection is made from those samples to cover one or more harmonics of the beat rate. In the present embodiment, the frequency range from 20 beats / minute to 500 beats / minute. Sample is selected. Change this value as needed to get heart rate harmonics Can also. Therefore, the output of the selection module is 256 samples. This embodiment State, the sample points 2-68 of the output of the FFT are used for further processing .   In the first pass of the process, the outputs from the selection modules 653, 655 are Provided to the magnitude module 656 and the red light magnitude module 658 Provided. The magnitude modules 656 and 658 have a magnitude function. Run, where the magnitude of each complex FFT point is Selected for each of the respective channels. Magnitude module 65 6 and 658 are output from the infrared light threshold module 660 and the red light threshold module. An input is provided to module 662.   The threshold modules 660, 662 examine the sample points point by point. To select points where the magnitude of each point exceeds a particular threshold. This Threshold is the highest detected among all remaining points in the snapshot. Set as a percentage of large size. In the present embodiment, the threshold operation The percentage of the solution is selected as 1% of the maximum magnitude.   After thresholding, the data points are sent to the point-by-point ratio module 670 Is done. The point-by-point ratio module calculates the ratio of red light value to infrared light value for each point. You. However, further testing has been done to qualify the ratio points. Will be As seen in FIG. 25B, suns from selection modules 653, 655 The pull point output also includes an infrared light phase module 690 and a red light phase Provided to module 692. Fez modules 690 and 692 are FFT Select the phase value from the job. The outputs of the phase modules 690, 692 are: Provided to the phase identification module.   The phase identification module 694 is paired with the phase modules 690 and 692. Calculate the phase difference between corresponding data points. Any two corresponding The phase difference between the points is determined by a specific threshold value (for example, If less than 0.1 radian, the sample point is eligible. Two corresponding If the sample points are too far apart in phase, Int is not used. The output of the phase threshold module 696 is RED / Provides a possible input to the IR ratio module 670. Therefore, the sample point To determine the ratio of a particular pair of, three tests are performed:       6. Red light sample must pass red light threshold 660       7. Infrared sample must pass infrared threshold 662       8. The phase between two points is at the phase threshold 696 Must be below a predetermined threshold determined   For eligible sample points, a ratio is taken in the ratio module 670. For points that do not qualify, the saturation is zero at the output of the saturation equation 672. Set.   The obtained ratio is obtained by using the saturation expression modules 502 and 5 in the statistics module 504. 20 is provided for the saturation equation module. In other words, the saturation equation Rule 672 receives the point-by-point ratio and pairs corresponding to discrete ratio points. Provide the saturation value as output. Saturation points from the saturation equation module 672 Output is a series of saturations that can be plotted as saturation versus frequency. Provide degree points. The frequency reference is in the complex FFT stage You can get points.   In one of two methods according to the invention, the selected arterial saturation module 6 As shown at 80, arterial (and venous) saturation can be selected. 1 According to one method, the arterial saturation value is calculated by the saturation module Point corresponding to the maximum saturation value for all point outputs from And can simply be selected. Alternatively, at different frequencies (points) The number of saturation values for each particular saturation value to form a histogram of the number of occurrences for each particular saturation value. Generate a histogram similar to the histogram of FIG. You can also. In each case, the arterial saturation can be obtained and the arterial saturation can be calculated. Provided as an output to the selected arterial saturation module at force line 682 Can be. To obtain venous saturation, do not use the maximum arterial saturation value, but zero. The minimum arterial saturation value at the point indicating the higher value is selected. Display 33 6 can be provided.   Using the fast saturation transform information, the clear plethysmograph shown in FIG. 25C. Waveforms and pulse rates can also be provided. Pulse rate and clear plethysmograph waveform In order to get, some additional features are needed. As shown in FIG. 25C Thus, the pulse rate and the clear plethysmograph waveforms are 00, spectrum analysis module 702, and window function module 704 Is determined using   As shown in FIG. 25C, the input of the window function module 700 is From the output of the FFT module 652 or 654. This embodiment Then, only one measurement signal is needed. Another of the window function module 700 The input is the arterial saturation obtained from the output of the selected arterial saturation module 680 .   The window function module is used to calculate the saturation values that are very close to the arterial saturation value. Implement a window function that is selected to pass frequencies that are substantially correlated to wavenumber. Run. In the present embodiment, the following window functions are selected. Where SAT_nIs the saturation value corresponding to each specific frequency for the sample point. Equal, SAT_artIs selected at the output of the selected arterial saturation module 680. Shows arterial saturation. This window function is used for the red light signal or infrared light signal. Applied to a window function input representing any complex FFT. Wi The output of the window function module 700 is the frequency spectrum determined by the FFT. A red light signal or an infrared light signal expressed by the Facts have been removed by the window function. Many possible window functions Should be understood. Furthermore, when using the above window function Understand that using higher power reduces more noise I want to be.   To get the pulse rate, the output point from window function module 700 is Provided to the spectrum analysis module 702. Spectrum analysis module 70 2 is the same as the spectrum analysis module 590 of FIG. In other words, The spectrum analysis module 702 uses the output point of the window function 700 By determining the first harmonic in the indicated frequency spectrum, the pulse rate To determine. The output of the spectrum analysis module 702 is the pulse rate.   Output of window function 700 inverted to get a clear plethysmograph waveform Provided to the window function module 704. Invert window function module Reference numeral 704 denotes the Kaiser of the window function module 648 or 650 in FIG. 25B. -Invert window function. In other words, the inverted window function 704 is Perform point-by-point inversion of the Kaiser function for the defined points.   Therefore, by using the complex FFT and the window function, Artery saturation, pulse rate and clear plethys by suppressing noise from the chismographic waveform A morphographic waveform can be obtained. The above description is mainly for operations in the frequency domain. Related to the operation, but with similar results in the time domain. It should be understood that the operation can be performed.Relation to generalized expressions   The measurements described for pulse oximetry above relate to a more general matter. each The signals transmitted through the finger 310 at the wavelengths λa and λb (log conversion) are shown as follows. Is:   The above variables can be best understood by referring to FIG. 6c as follows: A of 6c_ThreeAnd A_FourA layer containing venous blood in the test medium,_ThreeIs deoxygenated Hemoglobin (Hb),_FourIs oxygenated hemoglobin in venous blood (HB02). Similarly, FIG._FiveAnd A₆The layer containing Showing middle arterial blood, A_FiveIndicates deoxygenated hemoglobin (Hb);₆But Suppose we show oxygenated hemoglobin (HB02) in arterial blood. Therefore , C^vHb02 indicates the concentration of oxygenated hemoglobin in the vein, c^vHb is venous blood Indicates the concentration of deoxygenated hemoglobin in^vIs the thickness of the venous blood (for example, A_ThreeAnd A_Four(Thickness of a layer containing). Similarly, c^AHb02 is acid in arterial blood Indicates the concentration of denatured hemoglobin, c^AHb is deoxygenated hemoglobin in arterial blood Indicates the concentration of x^AIs the thickness of the arterial blood (eg, A_FiveAnd A₆Thickness of the layer containing You.   The wavelength selected is one in the visible red light range, ie, λa, and the infrared light range. , Typically λb. The typical wavelength value chosen is λa = 660 nm and λb = 910 nm. According to the constant saturation method, c^A _Hb02(T ) / C^A _Hb(T) = constant₁(Constant 1) and c^v _Hb02(T) / c^vH_b( t) = constant_Two(Constant 2). Acid in arterial and venous blood Elementary saturation changes slowly with sample rate, and this assumption holds. formula The proportional coefficients of (105) and (106) can be written as: In pulse oximetry, both equations (108) and (109) can be satisfied simultaneously.   r_a(T) multiplied by equation (106) and then subtracting equation (106) from equation (105) yields zero The secondary reference signal n ′ (t) is determined as follows:   r_v(T) multiplied by equation (106) and then subtracting equation (106) from equation (105) yields zero The primary reference signal s' (t) is determined as follows: s' (t) = S_λa(T) -r_v(T) S_λb(T) (110b)        = S_λa(T) -r_v(T) s_λb(T) (111b)   Under the constant saturation assumption, the primary signal portion s_λa(T) and s_λbAbsorption with (t) The contribution of the vein to the vein is not canceled. Therefore, the veins when the patient is at rest Low frequency modulated absorption by absorption and modulated absorption by vein absorption when the patient is moving Are indicated by the secondary reference signal n '(t). Therefore, the correlation scan Concomitant modification by finger venous blood under operation by Serra or other methods described above The absorption of both the adjusted absorption and the constant low frequency periodic absorption of venous blood is removed or Derived.   To illustrate the operation of the oxygen measurement of FIG. 11 to obtain a clear waveform, 26 and 27 show the input to the reference processor of the present invention using the constant saturation method. The signal measured for_λa(T) = S_λred(T) and S_λb(T) = S_λIR(T) is shown. The first segments 26a and 27a of each signal are Relatively less disturbed by facts; that is, the patient measures time in these segments While it did not work practically. Therefore, these segments 26a and 27a are large. In the body, the primary plethysmograph waveform at each measurement wavelength is shown. Each letter The second segments 26b and 27b of the signal are affected by motion artifacts. Patients were running during the time these segments were measured. This These segments 26b and 27b represent a large part of the measurement signal resulting from the operation. The displacement is shown. The third segment 26c and 27c of each signal is Is relatively unaffected by the 1 shows a primary plethysmograph waveform in FIG.   FIG. 28 shows the secondary reference signal n ′ (t) = measured by the reference processor of the present invention. n_λa(T) -r_an_λb(T) is shown. Again, the secondary reference signal n '(t) is Secondary signal part n_λa(T) and n_λb(T). Therefore, the secondary reference signal The first segment 28a of n '(t) is substantially flat, which is the first segment of each signal. It is said that almost no noise is generated by the operation in the elements 26a and 27a. It corresponds to the thing. The second segment 28b of the secondary reference signal n '(t) has a large variation. Indicates that the movement caused a large displacement in each measurement signal. It corresponds to. The third segment 28c of the noise reference signal n '(t) is substantially flat Which is also active in the third segment 26c and 27c of each measurement signal. Corresponds to the absence of artefacts. The reference processor is the primary reference signal s' (t) = s_λa(T) -r_vs_λbUnderstand that it can also be used to get (t) I want to be. The primary reference signal s' (t) generally exhibits a plethysmographic waveform.   FIG. 29 and FIG. 30 show the results obtained by the correlation canceller using the secondary reference signal n ′ (t). Defined primary signal s_λa(T) and s_λbApproximation s "to (t)_λa(T) and s "_λb(T) is shown. Note that the enlargement / reduction of FIG. 26 to FIG. Not the same to better show change. 29 and 30 show the reference processor. Shows the effect of the correlation canceller using the secondary reference signal n ′ (t) measured by ing. Segments 29b and 30b correspond to segments 26b and 27 of the measurement signal. It is not affected by noise generated by the operation as shown in FIG. In addition, the segment 29a, 30a, 29c and 30c are measurement signals free from noise due to operation. No change from signal segments 26a, 27a, 26c and 27c.   Secondary signal n estimated by correlation canceller using primary reference signal s' (t)_λa (T) and n_λbApproximation n ″ to (t)_λa(T) and n "_λb(T) is also the present invention It should be understood thatMethod of estimating primary signal part and secondary signal part of measurement signal in pulse oximetry   Implementing various embodiments of the above-described correlation canceller in software includes: It is relatively simple based on the above formula and the above description. However, constant saturation Method and a joint process estimator that performs joint process estimation using equations (54)-(64). Written in the C programming language to calculate the primary criterion s' (t) using step 572 A copy of the computer program subroutine is provided in Appendix B. This join The process estimator comprises a primary part and a secondary base, each correlated with the primary reference signal s' (t). In the primary signal part of the two measurement signals having a secondary part correlated to the quasi-signal n '(t) Estimate a valid approximation to This subroutine is specifically for pulse oximetry. Perform the steps shown in the flowchart of FIG. 9 for the monitor to be applied Another way. The two signals are measured at two different wavelengths λa and λb, and λ a is typically in the visible light region, and λb is typically in the infrared light region. An example For example, the invention specifically written to perform pulse oximetry using the constant saturation method In the clear embodiment, λa = 660 nm and λb = 940 nm.   For the variables defined by equations (54) to (64) in the description of the joint process estimation means The correspondence of the program variables is as follows: Δ_m(T) = nc [m]. Delta Γ_{f, m}(T) = nc [m]. fref Γ_b,m (t) = nc [m]. bref f_m(T) = nc [m]. ferr b_m(T) = nc [m]. barr F_m(T) = nc [m]. Fswsqr B_m(T) = nc [m]. Bswsqr γ_m(T) = nc [m]. Gamma ρ_{m, λa}(T) = nc [m]. Roh a ρ_{m, λb}(T) = nc [m]. Roh b e_{m, λa}(T) = nc [m]. err a e_{m, λb}(T) = nc [m]. err b κ_{m, λa}(T) = nc [m]. K a κ_{m, λb}(T) = nc [m]. K b   The first part of the program is the "initialization of the correlation canceller" of the operation block 120. Initialize registers 90, 92, 96 and 98 and intermediate variable values as shown by Run. The second part of the program is the operation block 130 "Left [Z^-1] Input of each delay element variable 110 as shown in Element variable 1 so that the value of Perform 10 time updates. The calculation of the degree of saturation is performed in a separate module. Various methods of calculating oxygen saturation are known to those skilled in the art. Such calculation is based on G. FIG. A. Mook et al. And Michael R. Newman (Michael R. Neuman) In the literature. Concentrations of oxygenated and deoxygenated hemoglobin Is determined, the saturation value is determined in the same manner as in equations (72) to (79), and the time t₁And t_TwoTo Measurements are different in time but close in time when saturation is relatively constant. Done. For pulse oximetry, time t = (t₁+ T_Two) / 2 The sum is determined by:   In the third part of the subroutine, the operation block 140 "Two measurement signal samples" Calculate the primary or secondary criterion (s '(t) or n' (t)) for , The proportionality constant r determined by the constant saturation method as in equation (3)_a (T) and r_vUsing (t), the signal S_λa(T) and S_λbPrimary group for (t) A quasi or secondary criterion is calculated. The saturation is calculated in another subroutine,_a (T) or r_vThe value of (t) is the composite measurement signal S_λa(T) and S_λb(T) Primary part s_λa(T) and s_λb(T) or secondary part n_λa(T) and n_λb( This subroutine is entered to estimate t).   As indicated by action block 150 "Zero Stage Update", The fourth part performs the update of the Z stage, and the backward prediction error b of the Z stage₀(T) is Set equal to the value of the reference signal n '(t) or s' (t) just calculated . Further, the intermediate variable F₀And B₀(T) (nc [m] .Fswsq in the program) r and nc [m]. Bswsqr) is calculated and the regression filters 80a and 80b Of the registers 90, 92, 96 and 98 in the least squares lattice type prediction means 70 at Used to set the value.   The fifth part of the program is an iterative loop, the loop counter M of which is shown in FIG. As indicated by the “m = 0” operation of the operation block 160, m = NC CELLS Reset to zero to have maximum value. NC CELLS is the loop counter This is a predetermined maximum value of the return. For example, NC Typical values for CELLS are 6-10 It is. The state of the loop is set so that the loop repeats at least 5 times, and the conversion Until the strike is satisfied or m = NC Continue iterating until CELLS. The conversion test is the sum of the weights of the four prediction errors plus the weight of the backward prediction errors. The total is a small number, typically less than 0.00001 (ie, F_m(T) + B_m(T ) ≦ 0.00001).   The sixth part of the program is the operation block 170 "mth LSL prediction means Stage order update ”, the forward and backward reflection coefficients Γ_{m, f}(T) And Γ_{m, b}(T) Register 90 and 92 values (nc [m] .fre in the program) f and nc [m], bref) are calculated. Forward and backward prediction error f_m(T ) And b_m(T) (nc [m] .ferr and nc [m] .be in the program rr) is calculated. Further, the intermediate variable F_m(T), B_m(T) and γ (t) (P Nc [m] in the program. Fswsqr, nc [m]. Bswsqr, nc [m ]. gamma) is calculated. The first cycle of the loop is the zero cycle of the program. Nc [0]. Calculated in the stage update part. Fswsqr and nc [0]. Bsw Use sqr.   The seventh part of the program in the loop started with the fifth part of the program Next to the m-th stage of the regression filter (s) at block 180 Regression coefficient registers 96 and 96 in both regression filters, as shown in And 98 values κ_{m, λa}(T) and κ_{m, λb}(T) (nc [m] .K in the program) a And nc [m]. K b) is calculated. Intermediate error signal e_{m, λa}(T) and e_{m, λb} (T) and variable ρ_{m, λa}(T) and ρ_{m, λb}(T) (nc [m]. err a and nc [m]. err b, nc [m]. roh a and nc [m ]. roh b) is also calculated.   The loop iterates until it passes the convergence test. Increasing the means of estimating the joint process The bundle test is also repeated each time the loop repeats until "run" Be executed. Weighted sum F of forward prediction error and backward prediction error_m(T) + B_m(T ) Is less than or equal to 0.00001, the loop ends. If not, The sixth and seventh parts of the program are repeated.   The output of this subroutine is the sample set S input to the program._λa( t) and S_λbA valid approximation s "for the primary signal part of (t)_λa(T) and s "_λb (T) or a valid approximation n ″ for the secondary signal_λa(T) and n "_λbAt (t) is there. Many sets of measurement signal samples have primary or secondary signal portions After an approximation to the combined process estimator has been estimated, the compilation of the output is For plethysmographic waves or motion artifacts at each wavelength λa and λb It provides a wave that is a valid approximation to it.   The subroutine in Appendix B is just one embodiment of implementing equations (54)-(64) Please understand that. The execution of the normalized QRD-LSL equation is simple, but the normalization equation Subroutine is attached as Appendix C, and subroutine of QRD-LSL algorithm is attached. The chin is attached as Appendix D.   The reference signal used in the correlation canceller such as the adaptive noise canceller is determined, and the A processor according to the invention for removing or deriving primary and secondary components from physical measurements One embodiment of a physiological monitor with a sensor has been described in the form of pulse oximetry. However, those skilled in the art will recognize that other types of physiological monitors can use the above described techniques. It will be obvious.   Further, using the signal processing technology described in the present invention, continuous time or substantially continuous The arterial and venous blood oxygen saturation of the physiological system Can also be calculated. These calculations are performed when the physiological system is intentionally affected. It can be performed whether or not it is turned on.   Furthermore, the primary or secondary signal part is removed to determine the reference signal Or the conversion and proportionality factor of the measurement signal other than the log conversion that allows it to be derived It will be appreciated that the determination of Furthermore, the proportional factor r is the first Although described in the text as a ratio of a portion of the signal to a portion of the second signal, the portion of the second signal A similar proportionality constant, determined as the ratio of a portion of the first signal to the first signal, is also provided by the processor of the present invention. Could be used. In the latter case, the secondary reference signal is generally n ′ (t) = n_λb( t) -rn_λaWill be similar to (t).   Furthermore, a correlation cancellation technique other than the joint process estimation is shared with the reference signal of the present invention. It is understood that it can be used for. These are, inter alia, least squares algorithms, Wavelet transform, spectrum estimation technology, neural network , Weiner and Kalman filters, but these Not limited.   One skilled in the art can use the techniques of the present invention for many different types of physiological monitors. You will understand that. Other types of physiological monitors include ECG, blood pressure monitors , Blood component monitor (other than oxygen saturation), capnogram, heart rate monitor, respiration monitor Or an anesthesia depth monitor, but is not limited thereto. In addition, Salizer, drug monitor, cholesterol monitor, glucose monitor, carbon dioxide Measuring the pressure and amount of substances in the human body such as elementary monitors or carbon monoxide monitors The techniques described above can also be used for monitors.   Further, the primary signal or the composite signal containing both the primary component and the secondary component The techniques described above for removing or deriving secondary signals are also very closely cross-correlated, Those skilled in the art will understand that the present invention can be applied to an electrocardiogram (ECG) signal derived from a given site. Will understand. A three-pole Laplacian electrode sensor as shown in FIG. , "IEEE Transactions on Biomedical Engineering", Vol. 39, No. 1, Bin He and Richard Jay in Issue 1 (November 1992). Ko “Body Surface Laplacian ECG Mapping” by Richard J. Cohen ECG sensor, which is a variant of the bipolar Laplacian electrode described in It should be understood that it can be used as a device. Used to meet the requirements of the present invention It should also be understood that there are a myriad of possible ECG sensor configurations to obtain. The same type of sensor can also be used for EEG and EMG measurements.   Further, those skilled in the art will recognize that a signal composed of reflected energy rather than transmitted energy. It will be appreciated that the techniques described above can also be implemented on the issue. Acoustic energy, X-ray Energy, gamma ray energy, or light energy (but not (Not limited) The primary or secondary part of the measurement signal of any type of energy It will also be appreciated that the minutes can be estimated by the techniques described above. Therefore, those skilled in the art The signal is transmitted through a part of the human body and reflected from the body as it returns to this part of the human body Understand that the technology of the present invention can be applied to monitors such as monitors using ultrasonic waves. Will. In addition, monitors such as echocardiography rely on transmission and reflection. The technique of the present invention can be used because it is very large.   Having described the invention with a physiological monitor, the signal processing techniques of the invention Used in many areas, including but not limited to the processing of biological signals. Can be. The present invention relates to a signal processor including a detector, comprising: A first signal portion including a second secondary signal portion and a second primary signal portion and a second secondary signal portion. In any situation of receiving a first signal including a second signal portion including a signal portion, An invention can be provided. Therefore, the signal processor of the present invention can Can be easily applied to the area.                                   Appendix A     Digital signal processing using an efficient multiphase regression all-pass filter   Fred Harris^*, Maximilian de 'Olle de Rantremanj And AG. Constantinides^{** *} United States 92182-0190 San Diego, California San Diego State University Department of Electrical and Computer Engineering^** United Kingdom London D.W. 7-2 Beatty Ixibi Signal Road Chemical Engineering Medical School Imperial University Faculty of Electrical Engineering Signal Processing Section   Abstract: Digital filters are significantly smaller computations as multiphase regression all-pass networks Can be realized at load. These networks are M-parallel all-pass subfields. As a sum of the phase shifts, constructively add the phase shift in the passband and add It is formed with a choice to be added destructively. Design we provide here The technology is gained by a new optimization algorithm (described in detail later) It is a prototype M-pass regression all-pass filter. These filters are Operated with a combination of over-coordinate transformation, re-sampling, and cascade selection. We follow the example of a sampling system obtained by the technology provided here Here are some designs using this new algorithm. International Conference on Signal P Reporting in processing             April-June, 1991, Florence, Italy 1. introduction   A standard finite impulse response (FIR) filter is tapped as shown in FIG. Can be modeled as a weighted sum of the contents of the delay line with loops. This combination The total is represented by (1), and the transfer function of this filter is represented by (2).   We assume that the frequency selectivity of the filter output is I noticed that it was due to phase shift between pulls. This phase sum is shown in FIG. At two different frequencies.   It is the phase shift that is related to the frequency-dependent gain of the filter. Please note that there is no. This weighting term mainly depends on the pass band width of the filter and Used to control stopband attenuation. Spatial beaming (with beam spoiler (1) and signal synthesis (peak-to-rms control) (2) The idea is to use phase shifts (at any amplitude) configured to serve the same purpose. It is to use.   The class of filter we describe here is that the weight of the FIR filter is frequency dependent. Replace with an all-pass sub-filter that shows a unit gain by existing phase displacement. Multi-phase structure As expected, the all-pass subfilter (shown in FIG. 34 (3))ⁿWith a first-order polynomial And M is the number of taps in the structure. This composite form uses a 41b multiphase structure. It is shown in FIG. 41a in a reflected modified form. The transfer function of this structure is (4 ).   Figures 44a and 44b show two and three full pass stages respectively for each pass. Derived from this new algorithm for two-pass and five-pass filters with 3 illustrates an example of an identical ripple design. Mutual suppression to optimal design (See the above journal Described) determines the possible number of stages per pass of the 5-pass filter by an optimal 5-pass filter. The overfilter uses three less stages than assigned to the design. Thus, the sequence (1, 1, 1, 1, 0), (2, 2, 2, 1, 2), (3, 3 , 2, 2, 2), etc. Similarly, the two-pass filter is a sequence (1, 1 ), (2, 1), (2, 2), (2, 3), (3, 3), etc. 2.2-pass filter   We consider that this structure, even if we narrow our point to a two-pass filter, It has been found that there are many interesting properties and clear uses. 2 as designed The pass filter is a 0.25f, 3dB bandwidth half bandwidth filter. This If the point is limited to half the sampling frequency, the filter is standard Is the same as the half-bandwidth Butterworth obtained by the bilinear transformation. This special In the case of a source, the real part of the root position is zero. If this zero point is the equivalent ripple stop Once optimized for stopband operation, the filter is a constrained elliptical filter become. This constraint is related to the characteristics of the complementary all-pass filter. We have two passes H for all-pass sections for filters₀(0) and H₁(0) (See FIG. 35) ), Half of the total and the difference A (0) and B (0) Determine the pass and high pass filters. We have all transit sections satisfy (5) From this (5), we can use the low-pass filter and high-pass filter shown in (6). To derive the power relationship between them.   We now interpret how this relationship affects the two-pass filter. Supplement For the foot filter we set the minimum passband gain to 1-ε₁Defined by the highest stop Band gain ε_TwoDefined by Replacing (6) with these gains gives (7) Is obtained.   ε₁Is small, then (ε₁)^TwoCan be neglected, and the equation (8) is derived. It is.   Therefore, if the stopband attenuation is selected to be 0.001 (60. dB), The band ripple is 0.0000005 (126.0 dB) or The minimum gain is 0.9999999 (-0.00000022 dB). Understanding As possible, these filters have a significantly flat passband. 2-pass equivalent The ripple stopband filter has a passband and stopband having -3.0 dB at 0.25f. This is an elliptic filter in which a stop band filter is a set. 2 pass and equivalent elliptical shape There are two major differences from the standard practice of Ruta. One is 4 to 1 for multiplication It is a saving, versus direct execution requiring eight coefficients (including scaling). However, the fifth-order half-bandwidth two-pass filter required only two coefficients. Second, all passes Structure shows unit gain in internal state, thus preserving internal state as in standard practice Does not require registers with extended precision.   As with any filter design, the transition bandwidth is outside the band attenuation at a fixed order. Can be swapped, or the transition bandwidth can be increased by increasing the filter order. And can be steeper for a fixed attenuation. By two-pass structure Mono for elliptical filters and Butterworth filters that can be implemented Although there is a graph, a simple approximate relationship of the equivalent ripple filter is expressed in (9). Where A (dB) is the attenuation in dB and Δf is the transition bandwidth Where N is the total number of all passing segments in the filter.                         2A. Hilbert transform filter   The Hilbert transform is considered to be a particular type of half-band filter. this Model and implement a wideband 90 'phase shift network. HT ( DSP application to form the analytic signal a (n) as represented in 10) This signal is often used in applications where this signal is restricted to the positive (or negative) frequency band. This is a signal having a spectrum.   A 2-pass filter can be formed in several ways as an HT, A simple method similar to the transform performed by the high-pass FIR filter (4) is Heterodyne half-band filter to sample frequency. Original filter If the response and the transformation are denoted by h (n) and H (Z) respectively, the heterodyne table The current can be easily understood from that of (11).   The conversion of the half-band filter to the HT filter is based on the transfer function of each Z¹To -jZ^-1Put on This is done by transposition. Since the polynomial of the all-pass subfilter is quadratic, Is transformed by changing the sign of the coefficients used in each all-pass subfilter. , And -j with low pass delay. These operations are shown in FIG.                     2B. Interpolation and decimation filters   Two-to-one resampling (up or down, generally interpolation and decimation Any half-bandwidth filter may be used to No. A finite impulse response (FIR) filter has a zero input point ( Upsampling) can be operated with multiphase distribution without processing, or Discarded output points (downsampling) can be calculated. this The option cannot be used for general regression filters, Optional for filters. The M-pass filter is one of the M-order sub-filters. Divided into multiphase segments due to the interaction of the partial delay and the delay line of each pass . This relationship is particularly simple to see in 2-pass filters. Index n₀ The impulse applied to the low side simultaneously contributes to the output via the high pass while the low side The excess delay Z from the band pass¹Does not produce output until the next index No. During that same next index, the high pass will^TwoRelated to polynomials Z¹Since there is no passage, it does not contribute to the output. Therefore, the filter is It gives successive samples of the impulse response from each other.   Downsampling is performed at each pass of the filter operating at a reduced input rate. By providing alternate input samples to. Second pass delay input As achieved by commutation, the physical delay of its passage is eliminated You. The filtered downsampled output is the sum of the responses of the two passes or Is generated by the difference. The calculation rate per input sample point is It is half of that of the unpulled filter, and therefore the two-pass filter of FIG. If the filter is used for downsampling, perform three multiplications per input point. Run.   Upsampling is done by reversing the downsampling process. Is done. This involves transmitting the same input to the two passes and converting between their outputs . Both forms of resampling are illustrated in FIG. As in the previous example 44a is used for upsampling, the output port Perform multiplications three times per int.   The cascading operation of supplementary output downsampling and upsampling is , Two-band maximally decimated quarter mirror filtering and reconstruction Execute (5).   Multiple stages of upsampling filter or downsampling filter Cascading means loading significantly less work per data point and Sample rate conversion can be achieved. For example, a cutoff frequency of 0.4f 1 to 16 up with 96 dB dynamic range for signals with numbers The sampler computes all of the orders (3,3), (2,2), (2,1) and (1,1). Requires a 2-pass filter with a continuously low number of pass stages. This is out With an average workload of about 2.7 operations per force point, for 16 outputs A total of 42 multiplications are required.                    2C. Iterative multiphase all-pass filter   The all-pass network mentioned earlier is ZⁿAnd especially 2-pass Z for the filter^TwoFormed by K times the spectrum of the original filter A useful transformation to obtain a higher order filter that indicates the iteration is^TwoTo Z^2kReplace with That is. For K = 2, the half-band filter concentrated on DC has DC and f, / It becomes a pair of quarter-band filters concentrated on 2. Iterative filter spectral repetition Is shown in FIG.   This conversion converts the spectrum of the prototype filter by 1 / K, Note that it reduces both over-bandwidth and transition width. Therefore, very steep The low transition polynomial of the 2K order obtained through the delay element This is achieved through the use of expressions. The energy in the repetitive spectral region can vary Filters incorporating order resampling and reduced order iterative filters Can be removed. 3.Effective all-pass filter bank   Effective spectral distribution cascades the output of the iterative low-pass and HT filters. Obtained by cading and resampling. For example, four quads The spectrum of the 4-channel filter bank concentrated on the runt is the complementary half-band filter. 39, which includes a resampler as represented on the left side of FIG. A pair of coupled HT filters follows. This 70 dB attenuation filter set The workload for this is three multiplications per output channel.   A 4-channel filter bank straddling the previous set (ie, four main Axially concentrated) by two complementary half-band filters, then a complementary half-band Iterated and formed by the sample filter and the resample HT filter be able to. This spectrum of this filter set is represented on the right side of FIG. You. The workload for this 70 dB attenuation filter set is Multiplication twice. 4.Other all-pass transformations   The prototype complementary all-pass filter set also provides a standard all-pass conversion (6). Therefore, it can be converted to a filter having an arbitrary bandwidth and an arbitrary center frequency. The low-pass transform for the half-band filter is shown in (12).   This conversion is based on the fact that the -3 dB point has₀So that the frequency of the filter is Warp a number variable. θ₀Is not π / 2, the second order of the all-pass subfilter The polynomial is z^-1Get the coefficient of the term. The synthesis filter operates twice per pole-zero pair. Multiplication is required, but this is still half the workload of the conventional baseline configuration. In addition, the pole at the origin converts to a first order all-pass filter (due to the delay of the second pass), Converting an inactive pole to an active pole requires an additional primary stage. FIG. Shows a warped version of the frequency of the filter shown in FIG. 44a. This file Both the filter and its complement can be obtained from the warped structure.   The band-pass conversion is represented by (13). In this conversion, the center frequency is frequency θ₁The frequency variable of the filter so that Warp. θ₁Is equal to 0.0, this transformation is C. Stated in the anti Equivalent to the inverse transformation. At any other frequency, this transform will add a second order all-pass filter to the fourth order Convert to a passing structure. This forms the four pole-zero pairs of the synthesis filter Requires four multiplications, which is still half the traditional standard form workload . In addition, the transformation involves a second-order all-pass step through the real poles (which are originally the image of the poles at the origin). To a pair of complex poles that requires a page. FIG. 45 shows the filter shown in FIG. Provides a frequency warped version. As before, this filter and its Can be obtained from the warped structure.   Other transformations and geometries (7) can be used with digital all-pass structures. Readers are directed to the references in this paper listed for a wealth of options .   (3) root, a_nThe Mth root of, and its reciprocal, are shown in FIG. As shown, they are arranged at equal intervals around the origin. In all passing stages Note that we realize M poles and M zeros per multiplication. This is for example The standard (factored or unfactored) standard form has one pole (or Is zero), whereas a 2-pass filter provides two poles and two zeros per multiplication Means that   Surrounding the unit circle, the all-pass sub-filter has a position near each pole-zero pair. It is clear that the phase shows a rapid change. Proper selection of pole positions By doing so, the phase displacement of each leg of the M-pass filter is To be matched or changed by a multiplication of 2π / M over the Torr interval it can. This is shown in FIGS. 43A-43D with M = 2 and M = 4. All pass The coefficients of the sub-filter are determined by the standard algorithm (3) or by Harris and Druer. Can be determined by new algorithms developed and recently published In a journal paper (SignalProcessing magazine). 5.Conclusion   The formation and usability of the all-pass polyphase filter structure was verified. Next we go through M-pass Several all-pass transforms are provided that can be easily applied to the pass-filter set. 2-through Although emphasized on over-networks, this paper can be easily extended to any M You. Mentions a new set of algorithms for designing M-pass filters But this will be reported in SignalProcessing in the near future. Would. We use this algorithm to generate the examples presented in this paper. Has shown the possibility of rhythm.   The main advantage of these structures is that they have very low power for a given filter task. Claude. Other papers (8, 9, 10) are presented by these filters The low sensitivity to the finite operation was discussed. The primary use of these filters Obstacles are due to these novelties and the lack of available design methods for calculating filter weights. It is like. This paper (in line with the excellent survey heard in the book) The next paper deals with the second issue. 6.Acknowledgment     This study was supported by the San Diego State University (SDSU) and the University of California Industry / Integrated Circuit System (ICAS) at NSD (UCSD) The University Joint Research Center (I / UCRC) was partially sponsored.                               Appendix B                             Appendix C                             Appendix D

────────────────────────────────────────────────── ─── Continuation of front page    (81) Designated countries EP (AT, BE, CH, DE, DK, ES, FR, GB, GR, IE, IT, LU, M C, NL, PT, SE), OA (BF, BJ, CF, CG , CI, CM, GA, GN, ML, MR, NE, SN, TD, TG), AP (KE, MW, SD, SZ, UG), AL, AM, AT, AU, BB, BG, BR, BY, C A, CH, CN, CZ, DE, DK, EE, ES, FI , GB, GE, HU, IS, JP, KE, KG, KP, KR, KZ, LK, LR, LT, LU, LV, MD, M G, MK, MN, MW, MX, NO, NZ, PL, PT , RO, RU, SD, SE, SG, SI, SK, TJ, TM, TT, UA, UG, UZ, VN (72) Inventors Elfadel, Ibrahim M.             United States 92677 California             Province Laguna Niguel Tessir Stori             To 30041 Apartments 218 (72) Inventor McKersey, Rex John             United States 92691 California             Mission Viejo Buendia Province               22832 (72) Inventor Webber, Walter Michael             United States 90046 California             Los Angeles Hillside Aveni             No. 108 7260 (72) Inventor Smith, Robert A.             United States 91720 California             Corona West Monterrey Row             De 415

Claims (1)

[Claims] 1. At least two measurement signals S₁And S_TwoIn a signal processor that processes Each measurement signal includes a primary signal portion s and a secondary signal portion n, and the signal S₁And S_TwoBut , S₁= S₁+ N₁And S_Two= S_Two+ N_TwoAccording to the relationship, s₁And s_Two, N₁And n_TwoIs s₁ = R_as_Two, N₁= R_vn_TwoHas the relationship of r_aAnd r_vIs a coefficient, and the signal At Sessa,   s₁And n₁Coefficient r that minimizes the correlation of_aDetermining the value of   The r_aCalculating blood oxygen saturation from the value of   Displaying blood oxygen saturation on a display; A method that includes 2. At least two measurement signals S₁And S_TwoIn a signal processor that processes Each measurement signal includes a primary signal portion s and a secondary signal portion n, and the signal S₁And S_TwoBut , S₁= S₁+ N₁And S_Two= S_Two+ N_TwoAccording to the relationship, s₁And s_Two, N₁And n_TwoIs s₁ = R_as_Two, N₁= R_vn_TwoHas the relationship of r_aAnd r_vIs a coefficient, and the signal At Sessa,   s₁And n₁Coefficient r that minimizes the correlation of_aDetermining the value of   Determined r_aUsing at least the first measurement signal or the second measurement signal Processing at least one of the first signal and the second signal to reduce n from one Forming a clean signal, A method that includes 3. Display the resulting clean signal on the display. 3. The method of claim 2, further comprising: 4. The first measurement signal and the second measurement signal are physiological signals, and And processing the physiological signal from the first and second measurement signals. 3. The method of claim 2, comprising determining data. 5. The physiologic parameter is arterial oxygen saturation. 4. The method according to 4. 6. The method according to claim 4, wherein the physiological parameter is an ECG signal. The described method. 7. A first part of the measurement signal indicates a heart rate plethysmograph and calculates a pulse rate The method of claim 2, further comprising the step of: 8. A physiological monitor,   Primary part s₁And the secondary part n₁Measurement signal S having₁Configured to receive A first input to be   Primary part s_TwoAnd the secondary part n_TwoMeasurement signal S having_TwoConfigured to receive A second input to be The first measurement signal S₁And the second measurement signal S_TwoBut,   S₁= S₁+ N₁   S_Two= S_Two+ N_Two According to the relationship s₁And s_Two, N₁And n_TwoBut,     s₁= R_as_Two, And n₁= R_vn_Two Have the relationship r_aAnd r_vIs the coefficient, The physiological monitor further comprises:   A scan reference processor, the scan reference processor comprising a plurality of r._aAccording to the possible values of Answer r_aIs multiplied by each possible value of the second measurement signal, and for each of the obtained values Subtracting the obtained value from the first measurement signal to provide a plurality of output signals;   A first input configured to receive a first measurement signal; A second input configured to receive a plurality of output signals from the processor. A decorrelation device, wherein the decorrelation device correlates the plurality of output signals with the first measurement signal. Provide multiple output vectors corresponding to   An input configured to receive a plurality of output vectors from the decorrelator; Integrating means having a plurality of output vectors in response to the plurality of output vectors. Determine the power corresponding to the   Extremum detection means having an input connected to the output of the integration means, Detect a power selected in response to the power corresponding to each output vector. , A physiological monitor characterized by the following: 9. Wherein the plurality of possible values correspond to a plurality of possible values of the selected blood component. The physiological monitor according to claim 8, characterized in that: 10. Wherein the selected blood component is arterial blood oxygen saturation. A physiological monitor according to claim 9. 11. Wherein the selected blood component is venous blood oxygen saturation. A physiological monitor according to claim 9. 12. 10. The method of claim 9, wherein the selected blood component is carbon monoxide. A physiological monitor according to claim 1. 13. 9. The method of claim 8, wherein the plurality of possible values correspond to physiological concentrations. A physiological monitor according to claim 1. 14． A physiological monitor,   Primary part s₁And the secondary part n₁Measurement signal S having₁Configured to receive A first input to be   Primary part s_TwoAnd the secondary part n_TwoMeasurement signal S having_TwoConfigured to receive A second input to be The first measurement signal S₁And the second measurement signal S_TwoBut,   S₁= S₁+ N₁   S_Two= S_Two+ N_Two According to the relationship s₁And s_Two, N₁And n_TwoBut,   s₁= R_as_Two, And n₁= R_vn_Two Have the relationship r_aAnd r_vIs the coefficient, The physiological monitor further comprises:   A conversion module, wherein the conversion module is configured to control the first measurement signal and the second measurement signal. Measurement signal and r_aGenerate at least one power curve in response to a plurality of possible values of Force   An extremum calculation module, wherein the extremum calculation module comprises the at least one R that minimizes the correlation between s and n in response to the power curve_aAnd choose this r_aof Calculate the corresponding saturation value from the values as output,   A display module, wherein the display module comprises the calculated Displaying the saturation value in response to a saturation output; A physiological monitor characterized by the following: 15. A signal processor for processing at least two measurement signals, wherein each measurement signal is The signal includes a primary signal portion and a secondary signal portion, and the first signal and the second signal are substantially located. According to a constant signal model, in the signal processor:   Sampling the first signal and the second signal over a predetermined period, and A first plurality of data points representing the first signal over a predetermined period of time. And obtaining a second series of data points indicative of said second signal. ,   Transforming the first series of data points to at least a frequency component and an amplitude component; Into a first series of transformation points having Transform to a second series of transformation points having at least frequency and amplitude components Including the step of   The first series of conversion points and the second series of conversion points are compared and shaken. Obtaining a third series of comparison values having a width component and at least a frequency component. See   At least one of the comparison values having an amplitude within a selected threshold value Selecting a comparison value,   A result value that matches a predetermined signal model from at least one selected comparison value Including the step of selecting A method comprising: 16. Determining a ratio of a first series of transformation points to a second series of transformation points; Selecting at least one of the comparison values is lower than the determined ratio. The method of claim 15, comprising selecting a ratio. 17． The step of determining a result value comprises calculating blood oxygen saturation from the selected ratio. 17. The method of claim 16, comprising: 18. 16. The method of claim 15, wherein the result value is blood oxygen saturation. the method of. 19. The method of claim 15, wherein the result value is a pulse rate. 20. A signal processor for processing at least two measurement signals, wherein each measurement signal is The signal includes a primary signal portion and a secondary signal portion, and the first signal and the second signal are blood components. Substantially according to the signal model for saturation, in the signal processor:   Sampling the first signal and the second signal over a period of time, A first plurality of data points representing said first signal over a period of time, Obtaining a second series of data points representing said second signal. See   Transform the first series of data points from time domain to frequency domain Obtaining a first series of transformation points and a second series of transformation points Wherein the first and second series of transformation points comprise an amplitude component and at least a frequency Component and   The frequency of the first series of transformation points versus the circumference of the second series of transformation points Determining a ratio of a series of amplitudes to a series of ratios of wave numbers;   From the ratio of the series of amplitudes, at least one having an amplitude within a selected threshold. Selecting two ratios,   Determining, from the at least one selected ratio, a result value that matches the signal model Including the step of A method comprising: 21. Wherein the ratio corresponds to blood oxygen saturation, and wherein the at least one ratio is selected. Step selects at least one ratio corresponding to a higher value of blood oxygen saturation 21. The method of claim 20, comprising: 22. Determining a result value comprises determining the blood acid from at least one selected ratio. The method of claim 21, comprising calculating elementary saturation. 23. Using a window function, the first series of transformation points or the Combining at least one of a series of two transformation points with said result value. Including   Performing a spectral analysis on the combined to obtain a pulse rate. Mm 23. The method of claim 22, wherein: 25. Using a window function, the first series of transformation points or the Combining at least one of a series of two transformation points with said result value Including   Performing an inverse window function to obtain a plethysmograph. 23. The method of claim 22, wherein: 24. 23. The method according to claim 22, wherein the result value is blood oxygen saturation. the method of. 26. A signal processor for processing at least two measurement signals, wherein each measurement signal is The signal includes a primary signal portion and a secondary signal portion, wherein the first signal and the second signal are substantially In accordance with the signal model, in the signal processor,   Sampling the first signal and the second signal over a predetermined period, A first plurality of data points representing the first signal over a period and the predetermined period And a second series of data points representing said second signal over Including   Performing a fast saturation transform using the first and second series of data points; Obtaining a series of transformed data points in the frequency domain,   Determining a saturation value selected from the plurality of saturation values, A method comprising: 27. Wherein the selected saturation value is arterial blood oxygen saturation. The method of claim 26. 28. Wherein the selected saturation value is venous blood oxygen saturation. The method of claim 26. 29. The step of performing the fast saturation conversion comprises the step of executing the first series of data points. A first plurality of intermediate transformation points and a second series of data points. Calculating a plurality of intermediate transformation points, wherein the method further comprises: Determining a pulse rate from the selected saturation value and the first plurality of intermediate conversion points; 27. The method of claim 26, comprising the step of: 30. A low-noise radiating element driver,   Including radiating elements,   A radiating element current source having a control input and an output, wherein the output is Provide current drive for the element,   A radiating element current source input having an output connected to a control input of the radiating element current source; A power switch, wherein the input switch inputs to one of at least two inputs Is connected   An output is connected to the radiating element, and an input is provided to one of at least two inputs. A radiating element current source output switch to be connected,   A radiating element switch control latch, wherein the control latch is connected to the radiating element current source input; A radiating element switch connected to a switch and said radiating element current source output switch. A control latch controls which input to the switch is selected; A low-noise radiating element driver, characterized in that: 31. Wherein said at least two inputs of said radiating element current input switch comprise said current 31. The power supply of claim 30, wherein the power supply is a power source and a ground. Noise radiating element driver. 32. A combination   A first signal propagating along a first propagation path and a second signal propagating along a second propagation path A detector responsive to the signal to provide a representation of the first signal and the second signal at an output. The first propagation path and a part of the second propagation path are located in the same propagation medium, The representation of the first signal at the output has a primary signal portion and a secondary signal portion, and the first signal The primary signal portion of the signal is attenuated along substantially the entire first propagation path and at the output A representation of the second signal has a primary signal portion and a secondary signal portion, wherein the primary signal portion of the second signal Attenuated along substantially the entire second propagation path,   A first signal processor having an input connected to the detector; A first signal and a second signal in response to the representation of the first signal and the second signal from the detector. Combining the second signal to form a first signal portion and a second signal portion of the first signal and the second signal; Either the primary reference signal or the secondary reference signal, which is a function of either Generate, A combination characterized by the following. 33. Responsive to the secondary reference signal and the representation of the first signal, the primary signal of the first signal; A second signal processor for deriving an output signal that is a function of the signal portion. 33. The combination according to claim 32, wherein 34. The second signal processor includes a decorrelator. 33. The combination according to 33. 35. The second signal processor is an adaptive noise canceling device. A combination according to claim 33. 36. The adaptive noise canceling device includes a joint process estimating means. 36. The combination according to claim 35. 37. The combined process estimating means comprises a least squares grid type predicting means and a regression filter; 37. The combination of claim 36, comprising: 38. Responsive to the primary reference signal and the representation of the first signal, Including a second signal processor for deriving an output signal that is a function of the signal portion. 33. The combination according to claim 32, wherein the combination comprises: 39. Responsive to the secondary reference signal and the representation of the first signal, Including a second signal processor that derives an output signal that is a function of the primary signal portion. 33. The combination of claim 32, wherein: 40. Responsive to the secondary reference signal and the representation of the first signal, Second signal processor for deriving an output signal having a component that is a function of a secondary signal portion 33. The combination of claim 32, comprising: 41. The detector detects a physiological function indicated by the first signal and the second signal 33. The combination of claim 32, wherein the combination is configured to: . 42. Wherein said detector is used to measure a blood component. A combination according to claim 41. 43. The blood component measured by the detector is blood gas. 43. The combination of claim 42. 44. The detector comprises a sensor responsive to electromagnetic energy. 42. The combination according to claim 41. 45. A first signal and a second signal received by the detector means through the propagation medium. Connected to the detector means to measure the plethysmograph waveform depending on the signal Further comprising electromagnetic means, wherein the propagation medium includes tissue in a living body, 33. The combination according to claim 32. 46. A pulse oximeter connected to the detector, wherein the pulse oximeter is Physiology dependent on a first signal and a second signal received by the detector through the body Monitoring the target condition, wherein the propagation medium includes a tissue in a living body. A combination according to claim 32. 47. The first signal and the second signal received by the detector through the propagation medium Connected to the detector, configured to derive a physiological condition dependent on the signal. 33. The combination of claim 32, further comprising a blood pressure monitor that is operated. Option. 48. An electrocardiogram connected to the detector, wherein the electrocardiogram passes through the propagation medium. An electrocardiogram depending on the first and second signals received by the detector means. Used to determine a condition, wherein the propagation medium comprises tissue in vivo. 33. The combination according to claim 32, wherein 49. The electrocardiogram includes a three-pole sensor having three concentrically arranged electrodes 49. The combination of claim 48, wherein: 50. A device for indicating a component of a substance,   A first signal processor for receiving a first input and a second input, wherein the first input is A plurality of reference signals, wherein each of the plurality of reference signals is associated with a second input. And   Providing the second input and each of the plurality of reference signals to the first signal processor means; Including means to provide,   The signal processor for each of the plurality of reference signals provided to the signal processor means; Means for detecting an output signal of the first signal processor means, wherein the output signal is Indicate the quality components, An apparatus for indicating a component of a substance characterized by the above. 51. Wherein said first signal processor means is a decorrelation device. An apparatus for indicating the components of a substance according to claim 50. 52. Wherein the first signal processor means is an adaptive noise canceling device. 51. A device for indicating the components of a substance according to claim 50. 53. The adaptive noise canceling device includes a joint process estimating means. 53. An apparatus for indicating the components of a substance according to claim 52. 54. The combined process estimating means includes a least squares lattice type predicting means and a regression filter. 54. An apparatus for indicating a component of a substance according to claim 53. 55. A second signal processor for receiving each of the output signals of the first signal processor means; And second signal processor means for multiplying each of the output signals. Forming an accumulated output signal indicative of the integrated output signal. The apparatus of claim 50. 56. Said detector means comprising a pole of said accumulated output signal of said second signal processor means; 56. The method of claim 55, further comprising providing an indication of a component of the substance in response to the value. An apparatus showing the components of the described substances. 57. The detecting means detects a change in the accumulated output signal of the second signal processor means. The method of claim 55, further comprising providing an indication of a component of the substance in response to An apparatus showing the components of the described substances. 58. The second input provided to the first signal processor means is a first signal or a second signal; Is one of the second signals, wherein each of the first and second signals is an arterial signal. Part and another signal part indicating venous blood,   A second signal processor for receiving the first and second signals and a plurality of signal coefficients; Means, the output of the second signal processor means being connected to the first signal processor. Forming the first input provided to the second signal processor means and providing an output of the second signal processor means. Are the first component associated with the arterial signal portion of the first signal and the second signal; And a second component associated with another signal portion of the second signal. 50. An apparatus showing the components of the substance of 50. 59. The other signal portion of each of the first and second signals provides an indication of human respiration. 59. A device for indicating a component of a substance according to claim 58, comprising: 60. At least one of the plurality of signal coefficients received by the second signal processor; And one of the first and second signals of the second input to the first signal processor. 59. The composition of matter of claim 58 associated with an arterial signal portion. Device. 61. At least one of the plurality of signal coefficients received by the second signal processor; One of the first signal and the second signal of the second input to the first signal processor means. 59. The substance of claim 58, wherein the substance is related to another signal portion of the two signals. Device for indicating the components. 62. A second input provided to the first signal processor means is a first signal or One of the second signals, wherein each of the first signal and the second signal is an arterial signal part. Minute and another signal part indicating venous blood,   A second signal processor for receiving the first and second signals and a plurality of signal coefficients; Means, the output of said second signal processor means being connected to said first signal processor. Forming the first input provided to the second signal processor means and providing an output of the second signal processor means. Is a component related to the arterial signal portion of the first signal and the second signal, or the first signal and the second signal. Including components related to other signal portions of the two signals; 51. An apparatus for indicating a component of a substance according to claim 50. 63. Another signal portion of each of the first and second signals includes an indication of human respiration. 63. An apparatus for indicating a component of a substance according to claim 62. 64. The substance of claim 50, wherein the substance is human tissue. Device showing the components of 65. The detector means comprises a first signal processor means for indicating a component of the substance; The composition of claim 50 responsive to the power of an output signal. Showing device. 66. An apparatus for calculating an arterial signal and a vein signal in a tissue in a living body,   A first signal propagating along a first propagation path and a second signal propagating along a second propagation path A first propagation path and a second propagation path including a detector configured to receive a signal Are located in the propagation medium, wherein the first signal is an arterial signal portion indicative of arterial blood; Another signal portion indicative of venous blood, wherein the second signal is indicative of arterial blood. Part and another signal part indicating venous blood,   Signal processor means, the signal processor having an input connected to the detector. Responding to the first signal and the second signal in response to the movement of the first signal and the second signal. Having a significant component that is a function of the pulse signal portion or any of the other signal portions Generate a signal, An apparatus for calculating an arterial signal and a venous signal in a tissue in a living body. 67. Another signal portion of each of the first and second signals includes an indication of human respiration. 67. An arterial signal and a vein signal in a tissue in a living body according to claim 66, characterized in that: A device that calculates 68. An apparatus for calculating an arterial blood component value and a venous blood component value in a tissue in a living body. hand,   Signal processor means for receiving a first input and a second input, wherein the first input is , One of the plurality of reference signals, each of the plurality of reference signals being associated with a second output. And   Dividing the second input and each of the plurality of reference signals of the first input into the signal processor; Means for providing to the   The signal processor means for each of the plurality of reference signals at the first input; Means for detecting the power of the output signal, wherein the power is the arterial blood component value and the static Indicating a pulse blood component value, Calculating the arterial blood component value and the venous blood component value in the in-vivo tissue apparatus. 69. The arterial blood component value and the venous blood component value are respectively arterial blood and venous blood. 69. The dynamics in tissue of a living body according to claim 68, wherein the oxygen saturation of the liquid is used. A device that calculates a pulse blood component value and a venous blood component value. 70. A method for minimizing noise in a physiological monitor, comprising:   Detecting a first measurement signal having a portion indicative of the frequency of a human heart rate; ,   At least one having a portion at least partially indicative of oxygen saturation of human blood Detecting two second measurement signals;   Isolating the frequency from the first measurement signal;   Providing a step of providing the at least one second measurement signal to a tunable filter. And   Filtering the at least one second measurement signal using the tunable filter. A step of filtering. A method for minimizing noise in a physiological monitor comprising: